Rectilinear strain sensing framework for real time compensation of structural distortions in precision machinery

Precision machinery such as machine tools use heavy duty structural elements to provide high accuracy and repeatability. However, external forces and thermal effects can still cause significant errors. Machine builders put significant emphasis on good design for error avoidance and more recently, utilise software-based error compensation methods to further improve performance. For geometric errors which result from build tolerances, and which normally only vary slowly over time, most Numerical Control systems provide functionality for pre-calibrated error compensation. For compensation of thermal errors, temperature sensors are often used to provide data for a model which calculates the effect of the temperature field on the machine structure. Although this method is widely researched, there is often significant residual error due the time-variant non-linear relationship between temperature and the error between the tool and workpiece. This may be exacerbated by introducing multi-material structural elements to reduce weight. Direct measurement and combining temperature with direct measurement can enable more precise modelling but can add significant cost of additional sensors. In this research a direct measurement method is applied using a series of short-range, ultra-low-cost displacement sensors, exploiting a particular arrangement of compact slotted photo-microsensors. These are organised in a novel rectilinear framework to enable detection of the bending of a machine tool structure. The design provides high resolution measurement of strain over arbitrary lengths and cost-effective hardware for permanently embedding on a machine. The system was applied to the ram of a high precision 3-axis machine tool and used to compensate the thermal error caused by running the integrated high-speed spindle. The residual error was reduced from 10 to 4 µm, even reducing the magnitude of the effect of the chiller cycles.