Advanced process monitoring and analysis of machining

Abstract: Milling is a processing technology massively applied in the metal manufacturing industry. The continuous demand for higher productivity and product quality asks for better understanding and control of the machining process. A better understanding can be achieved through experimental measurement and theoretical simulations and modelling of the process and its resulting product. In particular process monitoring and control is desirable for automated control and optimization of the process and in turn of productivity and product quality. However, due to the high complexity of the process, in particular of the interaction system/process/product, the above goals have only partially been achieved today, to a limited, unsatisfactory extent. The here presented research addresses the above needs and their progress by developing and starting to implement and study a highly sophisticated sensor and analysis platform for machining with the objectives of (i) advanced measurement and analysis of the milling process and (ii) advanced process monitoring and control. For a certain highly advanced high speed machining centre a comprehensive concept has been developed for the integration different kinds of sensors to the system. In particular the complex concept takes into account sophisticated analysis, research and development through the cooperation between different experimental and theoretical methods. The present thesis presents the developed platform concept. Due to the complexity of this sensor and analysis platform and its early stage of development, the here presented research work focuses on the implementation and investigation of selected sensors and analysis methods. Paper I screens and discusses the possibilities of the applied machining centre for sensor integration, particularly aiming at process monitoring. From the large variety of relevant sensors, particularly promising turned out to be acoustic emission sensors, force sensors, accelerometers, temperature sensors and optical sensors. Their potential and limitations are discussed. The specific concept developed for the addressed high speed machining centre is described in Paper II. Force sensors measuring the spindle force components, flexible accelerometers measuring the vibrations at almost any desired spindle or product location and Laser Doppler Vibrometry (LDV) for non-contact measurement of vibrations turned out to be the most promising sensors, to be implemented and studied in more detail in the following. Moreover, the concept recommends the cooperation of the sensors with modelling, simulation and analysis at different levels. Experimentally, the simulation of a static load to the spindle or even oscillating load through a magnetic shaker was realised in order to simulate the processing conditions for measuring and characterizing the spindle behaviour. The LDV- measurement enables precise identification of spindle oscillations at different locations. A particular challenge is the identification of stable operating points at high rotational speeds through stability lobes. For this purpose the vibrations of a certain complex, thin walled aerospace structure were measured by LDV and in good agreement simulated by FEA for different machining process stages in order to derive stable lobe areas. Beside implementation and first testing of the above mentioned methods, the paper describes and discusses possible concepts for a closed loop control of the process. In Paper III the LDV method is for the first time and successfully applied for measuring the fee-running rotating spindle of a high speed machining centre. In particular the roundness of the dummy tool applied was measured at high accuracy during rotation, as verified by mechanical measurement. Moreover, the LDV permits the measurement of misalignment of other geometrical and kinematical imperfections of the spindle rotation. Paper IV describes in more detail the verification of FEA through LDV for the complex aerospace structure, presented in Paper II. Both, vibrations of the unconstrained and of the clamped workpiece are studied. Moreover, the workpiece vibration characteristics was studied at four different machining stages, starting from aluminium stock and ending in a thin walled structure after removal of 97% of the material. In particular the corresponding stability lobes were predicted by the aid of the FEA, compared with modal analysis and with selected milling experiments. When machining a monolithic structure that reduces it's volume to such an extent gives several problem: Clamping of a thin walled structure is hard due to its flexibility, The changing dynamics of the workpiece has the affect that prediction of the vibrational properties is very hard to realise and finally the milling is hard because of the fact that the walls of the structure bends when a force is applied from the milling tool. Paper V discusses the suitability of different sensors for process monitoring, by applying them on-line during machining experiments for steel and aluminium, each at three different feed rates. The signals of force sensors in three dimensions arranged below the workpiece, of accelerometers in two directions mounted at the spindle housings and of LDV spotting the tool fixture ring were acquired on-line. The six signals were analysed, extracting mean values that were compared to the achieved surface roughness for the six different milling conditions. From evaluation of the stability of the signal and its coherence with the surface roughness trends the suitability of each sensor was discussed. While the sensors lateral and vertical to the feeding direction were not suitable, the sensors measuring in feeding direction were most promising, in particular the accelerometer and the corresponding LDV. The context between the roughness, process and sensor signals is discussed. Summarizing, a sophisticated sensor and analysis platform concept was developed. Selective sensors and methods have been implemented and studied with the purpose of system characterization, process understanding and process monitoring. The promising results encourage for continuation of this research programme.