Advances in Research

Machine tool accuracy, productivity and dynamic stability are constrained less by the average stiffness of the frame than by the behaviour of a small number of comparatively compliant elements within it. Finite element (FE) analysis has become the standard instrument for locating such elements and for redesigning them, and the resulting body of work now spans beds, columns, crossbeams, rams, slide tables and spindle housings across turning, milling, grinding and gantry-type configurations. This review examines that literature critically rather than descriptively. It traces how structural weakness is defined operationally, how the choice of definition interacts with modelling assumptions, how optimisation problems are formulated and solved, and how the resulting designs are verified. Three linked problems recur. First, weakness is operationalised through mutually inconsistent measures, including static compliance contribution, modal strain energy share, flexibility-matrix magnitude and design-variable sensitivity, and cross-criterion comparison on a common structure is almost absent, so the identity of the weakest component is often criterion-dependent rather than physical. Second, the components identified as weak are strongly conditioned by assumptions about joints, supports, load cases and axis configuration; where interfaces are represented as bonded contact, compliance that physically resides in guideways, bearings and bolted flanges is displaced onto the castings that surround them. Third, reported performance gains are usually assessed against the same numerical model that generated them, and where experimental model updating is applied, the fitted interface parameters absorb model error in a manner that limits the independence of subsequent validation. Well-supported conclusions are distinguished from provisional ones: sensitivity-guided rib and wall-thickness redesign consistently improves the stiffness-to-mass ratio relative to intuition-based design, whereas claims of general superiority for particular topology optimisation variants remain weakly evidenced. Priorities are proposed for criterion benchmarking, interface-aware formulations, process-coupled objectives, configuration-robust optimisation and transparent reporting.