**1. Introduction**

In production, the aim is to achieve full automation of manufacturing through autonomous machining processes. In this context, process monitoring systems are an important part of modern production plants. They protect machines and machine operators from damage, reduce downtime, and improve workpiece quality by eliminating chatter [1]. Rehorn et al. estimated, if the computer numerical control (CNC) machine tool is equipped with a monitoring system, downtime can be reduced by up to 20% while productivity can be raised by 50%. Machine utilization even increases by more than 40% [2]. The increasing complexity of series production represents a challenge for process monitoring. This also includes the use of new design strategies, e.g., by combining different materials in hybrid workpieces. These workpieces with locally adapted properties offer a promising approach for designing components with energy-efficient applications while reducing the use of high-alloy materials. This is the main focus of the Collaborative Research Center (CRC) 1153, where material combinations are joined and formed with different processes and then machined [3,4]. Due to different material properties and chemical compositions, the machining properties and chip formation mechanisms change during the cutting process and pose challenges for machining. In particular, the force gradient during material transition can lead to increased tool wear and the various material properties affect the workpiece geometry and surface quality. In addition, the trend to single part production or small batch sizes as well as the high demands on surface quality and manufacturing tolerances pose further challenges. For this reason, modern process monitoring systems must be provided with process information of the highest possible quality.

In this context, more monitoring approaches that combine di fferent signals and features are being researched [5–7]. A process parameter that o ffers a high level of information regarding various production errors is the process force [2,8]. Balsamo et al. showed that a catastrophic tool failure during turning can be monitored by multi-sensor signal processing of force and acoustic emission [9]. Jie et al. successfully developed an approach for tool condition monitoring during the machining of titanium alloys, which is also based on the force and acoustic emission signal [10]. In addition, the development of a cloud-based framework for monitoring manufacturing processes for online process monitoring services takes into account process force information due to its high sensitivity [11]. In general, the machine tool does not provide detailed information about forces. Therefore, various approaches have been studied in the past to extract this information parallel to the machining. A method to obtain information about the process force is to extract the process components from control signals [12,13]. This is associated with a high modelling e ffort, which has to be carried out for each machine due to manufacturing and assembly tolerances. By using external sensors, process forces can be measured with high sensitivity [14,15]. These sensor systems require high acquisition costs, are limited in their flexibility, and restrict the installation space for workpieces [16]. Further approaches to determine process force information have focused on modelling forces by virtual processing [17]. The simulations reach their limits if tool wear and thermal e ffects are to be considered in real time. In addition, a high modeling e ffort is required to take the structural dynamic e ffects into account. To overcome these challenges, Denkena et al. developed the idea of the "feeling" machine tool. They integrated sensory machine components, which are placed close to the process and are located directly in the force flow. These sensory components are equipped with strain gauges that measure the structural strain caused by the process force. Forces are then reconstructed from the strains. With this approach, the static and dynamic behavior of the machine structure is generally maintained and a flexible force measurement is realized. In milling machines, the spindle slides have been modified to become a sensory machine component. Based on the force sensing properties of the slide, the static tool deflection was determined during the process. The workpiece quality was improved by online adaptation of the tool position using an axis o ffset [18]. The monitoring of geometrical errors of the workpiece was investigated by a feeling clamping system for a milling process. Therefore, holes were drilled into the workpiece to represent a material defect for a subsequent flank milling operation. A dynamometer was used as a reference. The errors could be detected by both measuring systems based on confidence limits. However, due to a lower signal-to-noise ratio, the feeling clamping system had a lower sensitivity to the error [19]. Denkena et al. integrated metallic strain gauges in four carriages of a linear guide of a spindle-driven position axis [20]. The load on the carriages could be measured in two directions orthogonally to the direction of travel. The quality of the recorded measurement signals was suitable for detecting parallelism errors in the linear guide. Based on the achieved force, sensitivity process monitoring could not be performed.

This paper focuses for the first time on the investigation of the suitability of a feeling machine for process monitoring of turning operations. In this context, the process force is measured based on a feeling turret during machining and is compared with the measured forces of a dynamometer. Individual process errors, which a ffect the process and workpiece quality, are investigated. These include tool breakage, tool wear, and the varying material transition position of hybrid workpieces. Based on the considered process errors and the measuring accuracy of the feeling turret, the sensitivity of the monitoring is discussed for various process parameters.

#### **2. Materials and Methods**
