M.F. Zaeh’s research while affiliated with Technical University of Munich and other places

This paper presents an efficient and industry-ready system architecture that enables both the control of machining operations and the high-frequency acquisition of controller data and external sensor signals. Using the recorded data, a dexel-based mechanistic cutting force model, which enables the estimation of cutting forces for complex tool geometries in arbitrary machining operations, is parameterized via an instantaneous cutting force identification method. In the introduced identification method, Bayesian Optimization and Dynamic Time Warping are combined to avoid time-consuming and error-prone synchronization of measurement and simulation. The suitability of this approach was demonstrated by performing cutting experiments at various radial depths of cut and feed rates per tooth. Thereby, a good agreement between simulation and measurement could be observed.

In the course of the digitization of modern production systems, a reliable parameterization of the digital twin of machining processes is essential. For example, the digital representation of milling operations enables the process parameter selection without time-consuming and expensive test series by using stability lobe diagrams (SLD). However, the parameterization of the underlying process force model with very few cutting force experiments can prevent a reliable process design, as errors in the parameterization process are propagated to the stability analysis. Therefore, a novel two-step methodology is proposed to provide probabilistic credible intervals for conventional stability lobe diagrams: First, the unknown parameters of the process force model are estimated using a Bayesian regression method. Secondly, the estimated probability distributions of the process force parameters are used to quantify the uncertainty of the stability boundary using a mechanistic process force model. The proposed methodology is particularly characterized by its low computational cost, since time-consuming and computationally expensive Monte Carlo procedures are avoided. Instead, the methodology relies on the analytical derivation of the model parameters’ posterior probability distribution and on polynomial chaos expansion (PCE) algorithms to quantify the uncertainty in the final stability analysis.

The comparison between measured and simulated machining forces enables the evaluation of workpiece quality, process stability, and tool wear condition. To compute the machining forces that occur, mechanistic cutting force models are typically used. The cutting force coefficients (CFCs) of mechanistic force models are directly linked to the mechanics of chip formation and, thus, depend on the tool-workpiece combination and on the prevailing cutting conditions. CFCs are usually identified via the average cutting force identification method, which requires the execution of cutting tests under defined test conditions. Hence, determining CFCs for different cutting conditions is time-consuming and expensive. In this paper, the performance of an instantaneous CFC identification approach based on Bayesian Optimization during the machining of arbitrary workpiece geometries is studied. Bayesian Optimization is well suited for global optimization problems with computationally expensive cost functions. The simulated cutting forces are calculated using a dexel-based cutter workpiece engagement simulation and the actual cutting forces are measured during the machining process using a dynamometer. Thus, an efficient identification of CFCs could be achieved.

Abstrac Chatter is the main limiting factor affecting the material removal rates of machine tools and is caused by the machine's most flexible structural mode shapes. The use of active vibration control systems can damp the structural mode shapes and, in turn, significantly increase the chatter-free depth of cut. In order to reduce the commissioning effort, previous publications have introduced methodologies to automatically tune various control strategies for active damping. This paper presents a new automatic tuning approach for a robust model-based controller, which uses particle swarm optimization to find the best-performing control parameters. The proposed method was tested on both a 5-axis milling machine and a vertical lathe, together with an automatically tuned direct velocity feedback controller and an adaptive controller. The performance was evaluated by conducting extensive cutting tests under industrial operating conditions. All controllers and automatic tuning methodologies led to notable increase in chatter-free material removal rates, reduced vibration amplitudes, and improved surface roughness.

A central element in the surgical treatment of bone fractures is the functional design of bone plates. A novel approach is the use of polyaxial locking screws, which allow the surgeon to respond to different surgical situations. Furthermore, additive manufacturing enables the production of patient-specific bone plates. By combining both technologies, the surgeon benefits from patient-specific implants and the flexibility to react to surgery events. This study evaluates the combination of the two approaches. For this purpose, test specimens that replicate the bolting of the locking mechanism were constructed. These test specimens were subjected to tensile tests in order to determine the maximum forces that the bolted connection can withstand. Conventional bar material and additively manufactured Ti-6Al-4V, which were processed through powder bed fusion using an electron beam (PBF-EB), were used as the base material for the test plates. The investigations showed that higher maximum forces could be achieved with the additively manufactured specimens.

State of the art machine tool controllers offer several Internet-of-Things (IoT) interfaces for machine data acquisition using industrial or edge computers. However, the available data exchange rates for these communication platforms are limited to a few hundred Hertz. As the data is not available in high frequency resolution, such a network communication is not suitable for monitoring and optimizing highly dynamic machining processes. This paper describes an efficient system architecture, which enables the acquisition of internal machine data as well as the high frequency sampling of external sensors. Based on this data, an Operational Modal Analysis (OMA) approach can be used to determine the tool tip dynamics during the machining process. Identification of tool tip frequency response requires the reconstruction of the excitation of the machine tool structure, i.e. the occurring machining forces. For this purpose, an approach relying on monitoring the commanded motor currents is applied.

A 3-D finite element model of rotary friction welding was developed using a viscoelastic Maxwell model. The thermal softening of the material was accounted for in the viscous part of the material model as well as in an additional friction model. The thermal and mechanical material data were adapted from existing data of tempering steel AISI 4140. The final model was used to predict rotary friction welding of shafts using three different parameter settings. The simulation results showed that the melt temperature was never exceeded at the faying surfaces, which is seen as a main characteristic of friction welding. Subsequent validation also demonstrated that the shortening of the specimen as well as the resulting flash geometry could be predicted successfully.

Friction stir welding is an advanced joining technology that is particularly suitable for aluminum alloys. Various studies have shown a significant dependence of the welding quality on the welding speed and the rotational speed of the tool. Frequently, an inappropriate setting of these parameters can be detected through an examination of the resulting surface defects, such as increased flash formation or surface galling. In this work, two different learning-based algorithms were applied to improve the surface topography of friction stir welds. For this purpose, the surface topographies of 262 welds, which were performed as part of ten studies, were evaluated offline. The aim was to use reinforcement learning and Bayesian optimization approaches to determine the most appropriate settings for the welding speed and the rotational speed of the tool. The optimization problem was solved using reinforcement learning, specifically value iteration. However, the value iteration algorithm was not efficient, since all actions and states had to be iterated over, i.e., each possible parameter combination had to be evaluated, to find the best policy. Instead, it was better to solve the optimization problem directly using the Bayesian optimization. Two approaches were applied: both an approach in which the information from the other studies was not used and an approach in which the information from the other studies was used. On average, both the Bayesian optimization approaches found suitable welding parameters significantly faster than a random search algorithm, and the latter approach improved the result even further compared with the former approach. Future research will aim to show that optimization of the surface topography also leads to an increase in the ultimate tensile strength.