Thomas Semm’s research while affiliated with Technical University of Munich and other places

Chatter vibrations originated by the machine structure are a major limitation for the productivity of ram based machines performing heavy duty operations. Consequently, the damping of the machine structure has a capital importance. It is known that interfaces and guideways are the main origin of damping. Recently, the use of active dampers has been introduced in industry. In this work, the damping of hydrostatic and rolling guideways with and without active damping has been experimentally identified and compared using receptance coupling. The results show that hydrostatic guidance can introduce 3–4 times more damping than a roller based system. However, the introduction of active damping is a game changer enhancing damping more than 30 times.

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.

Due to their critical influence on manufacturing accuracy, machine tool feed drives and the monitoring of their condition has been a research field of increasing interest for several years already. Accurate and reliable estimates of the current condition of the machine tool feed drive’s components ball screw drive (BSD) and linear guide shoes (LGSs) are expected to significantly enhance the maintainability of machine tools, which finally leads to economic benefits and smoother production. Therefore, many authors performed extensive experiments with different sensor signals, features and components. Most of those experiments were performed on simplified test benches in order to gain genuine and distinct insights into the correlations between the recorded sensor signals and the investigated fault modes. However, in order to build the bridge between real use cases and scientific findings, those investigations have to be transferred and performed on a more complex test bench, which is close to machine tools in operation. In this paper, a condition monitoring test cycle is developed for such a test bench. The developed test cycle enables the recording of a re-producible data basis, on which models for the condition monitoring of BSDs and LGSs can be based upon.

Models that are able to accurately predict the dynamic behavior of machine tools are crucial for a variety of applications ranging from machine tool design to process simulations. However, with increasing accuracy, the models tend to become increasingly complex, which can cause problems identifying the unknown parameters which the models are based on. In this paper, a method is presented that shows how parameter identification can be eased by systematically reducing the dimensionality of a given dynamic machine tool model. The approach presented is based on ranking the model's input parameters by means of a global sensitivity analysis. It is shown that the number of parameters, which need to be identified, can be drastically reduced with only limited impact on the model's fidelity. This is validated by means of model evaluation criteria and frequency response functions which show a mean conformity of 98.9 % with the full-scale reference model. The paper is concluded by a short demonstration on how to use the results from the global sensitivity analysis for parameter identification.

The dynamic behavior of a machine tool is the result of the mass, stiffness and damping distribution, which are highly influenced by local properties. As a result, it can change significantly for different axis positions. Modeling the dynamic behavior accurately and at the same time efficiently is challenging and prevents the holistic optimization of machine tools. Therefore, this publication presents an approach to identify optimization potentials using flexible multibody models with local damping information. The energy distribution as well as the receptance are analyzed for different machine states to determine damping-based optimization capabilities. A comparison to state-of-the-art models shows the improved efficiency of the presented procedure.

The conventional manufacturing of aircraft components is based on the machining from bulk material and the buy-to-fly ratio is high. This, in combination with the often low machinability of the materials in use, leads to high manufacturing costs. To reduce the production costs for these components, a process chain was developed, which consists of an additive manufacturing process and a machining process. To fully utilize the process chain’s capabilities, an integrated process planning approach is necessary. As a result, the work sequence can be optimized to achieve the economically most suitable sequence. In this paper, a method for a joint manufacturing cost calculation and subsequent decision-based cost minimization is proposed for the wire and arc additive manufacturing (WAAM) & milling process chain. Furthermore, the parameters’ influence on the results and the magnitude of their influence are determined. These results make it possible to design an economically optimal work sequence and to automate the process planning for this process chain.

Preliminary studies have shown the superiority of convolutional neural networks (CNNs) compared to other network architectures for determining the surface quality of friction stir welds. In this paper, CNNs were employed to detect cavities inside friction stir welds by evaluating inline measured process data. The aim was to determine whether CNNs are suitable for identifying surface defects exclusively, or if the approach is transferable to internal weld defects. For this purpose, 120 welds were produced and examined by ultrasonic testing, which was the basis for labeling the data as “good” or “defective.” Different types of artificial neural network were tested for predicting the placement of the welds into the defined classes. It was found that the way of labeling the data is significant for the accuracy achievable. When the complete welds were uniformly labeled as “good” or “defective,” an accuracy of 98.5% was achieved by a CNN, which was a significant improvement compared to the state of the art. When the welds were labeled segment-wise, an accuracy of 79.2% was obtained by using a CNN, showing that a segment-wise prediction of the cavities is also possible. The results confirm that CNNs are well suited for process monitoring in friction stir welding and their application enables the identification of various defect types.

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.

Der digitale Zwilling als Modell gewinnt sowohl für die Entwicklung neuer Maschinengenerationen als auch für Simulationen parallel zum Betrieb stark an Bedeutung. Zur Erstellung entsprechender Modelle sind moderne flexible Mehrkörpersimulationsprogramme besonders geeignet. Im Rahmen dieses Beitrags wird die Simulationsumgebung MORe präsentiert, die sich unter anderem durch ihre Benutzerfreundlichkeit und ihre Recheneffizienz auszeichnet. Zudem ist die Berücksichtigung von Effekten möglich, die bisher im industriellen Umfeld kaum betrachtet wurden, wie beispielsweise Dämpfung. The digital twin is becoming increasingly important for the development of new machine generations and for process parallel simulations. Modern flexible multi-body simulation programs are particularly suitable for creating the relevant models. In this paper, the simulation environment MORe is presented, which is characterized by its user-friendliness and its computational efficiency. Furthermore, it is possible to study effects such as damping, which have hardly been considered in industrial environments so far.

The estimation of remaining useful life (RUL) of machinery is a major task in prognostics and health management (PHM). Recently, prognostic performance has been enhanced significantly due to the application of deep learning (DL) models. However, only few authors assess the uncertainty of the applied DL models and therefore can state how certain the model is about the predicted RUL values. This is especially critical in applications, in which unplanned failures lead to high costs or even to human harm. Therefore, the determination of the uncertainty associated with the RUL estimate is important for the applicability of DL models in practice. In this article, Bayesian DL models, that naturally quantify uncertainty, were applied to the task of RUL estimation of simulated turbo fan engines. Inference is carried out via Hamiltonian Monte Carlo (HMC) and variational inference (VI). The experiments show, that the performance of Bayesian DL models is similar and in many cases even beneficial compared to classical DL models. Furthermore, an approach for utilizing the uncertainty information generated by Bayesian DL models is presented. The approach was applied and showed how to further enhance the predictive performance.