An automated Y-cutting system has been developed by the Automation Systems team of Industrial Research Limited. This robotic device performs the Y-cut operation on sheep carcasses. The robotic Y-cutting system must deal with a variety of carcass shapes and sizes, and it is important that process faults are detected, diagnosed and corrected as quickly as possible. This thesis addresses the fault detection and path optimisation requirements of the Y-cutting system. The development of a neural network-based fault detection module is documented. This module classifies process faults using axial motor current data from the Y-cutting robot. The module successfully classifies 98% of the presented cut signals during offline training, and 100% of cuts during an extended trial in an Australian meat-processing plant. An online training scheme is implemented to allow for the retraining of the neural network weights as required. The fault detection module is extended to handle a greater number of fault conditions and to detect variations in the process load. A path optimisation algorithm is developed to optimise the parameters that define the cut-path of the robot based on the output of the fault detection module. A line-search within the parameter space is used to estimate the position of the optimum parameter value. The optimisation of fifteen path parameters requires 4760 simulated Y-cuts, equating to approximately 1.5 days of processing in a typical meat-plant. This is significantly faster than the existing method for manually tuning the Y-cutting system. The fault detection and path optimisation systems can be generically applied to other robotic systems produced by Industrial Research Limited for the handling and processing of highly varying natural products.