Keywords

1 Introduction

Due to increasing demands in terms of energy efficiency and reduction of CO2 emissions, the metal processing industry is also subject to growing pressure. Because this branch is responsible for around 23% of Germany’s total annual emissions (in 2019, 183 m t CO2e), particular attention is being paid to this sector [1]. In addition, batch sizes are decreasing while the number of variants is increasing. In order to be able to continue to produce competitively, this requires flexible and at the same time energy-efficient production [2].

In this paper, two sections are considered to contribute to efficiency increase and pollutant reduction in the mentioned industrial sector. The first section analyzes in which form relevant data can be obtained during machining processes. In addition, the quality of the respective machine-internal process data is considered in order to allow meaningful conclusions to be drawn about the real machining process.

The second section focuses on the human being in the form of the machine operator in order to take optimal advantage of his experience and qualifications. For this purpose, the machine-internal data obtained is to be processed and visualized by means of an assistance system in such a way that it can be quickly perceived by the operator and decisions can be made independently to increase the efficiency of the machine and consequently of the process.

2 Measurement of Internal Machine Data

In order to be able to evaluate machining processes on the basis of machine-internal data, it is first necessary to consider which data can be called up for process optimization via the machine tool control and what consequently significance they offer. The use of machine-internal data is justified by the challenge that installed machine tools of existing production lines can only be retrofitted with external measuring sensors at high investment costs. This poses a major economic hurdle, especially for medium-sized companies, and is often considered too high financial risks. In addition, there is a risk that complex measurement setups at laboratory level, such as force plates based on piezo-electric elements, will negatively influence the actual machining process, falsify occurring process forces and limit the flexible use of the machine tool. Furthermore, a high level of experience and qualification is required to evaluate the data obtained and to implement measures for process optimization.

The general approach is therefore to evaluate and present the data provided in such a way that the findings can be used to increase efficiency and thus reduce CO2 emissions. Relevant machine-internal data can be drive currents of spindle motors, but also current positions, speed or acceleration values of individual drive axes. Depending on the control system, these are measured in the position control cycle at the respective frequency inverter of a drive, sent to the Numerical Control Unit (NCU) and finally output at the Human-Machine Interface (HMI) or external measuring computer.

2.1 Suitability of Internal Machine Data for Process Optimization

Because monitoring the drive power of spindle and feed drives to protect against overload is standard equipment on many machine tools, it is an obvious approach to check their suitability for process monitoring and optimization [3].

Own investigations showed that the evaluation of the torque-forming current consumption of the axis drives is not suitable for the evaluation of the loads arising in the process, such as cutting, feed and passive forces. On the test machine used, type GROB G350, it was noticed, that the holding currents caused by the raised machine table are many times higher than the current changes caused by machining. However, the situation is different when evaluating the drive power of the spindle drives. Although the basic load, which occurs when the spindle is rotating without machining, has to be compensated, current changes and thus torques resulting from the machining process can be clearly detected.

Fig. 1.
figure 1

Spindle load curve of typical roughing milling process

Full size image

Figure 1 shows the spindle load during a roughing milling process. In this case, the tool utilization in sector 2 is below the referencing maximum from sector 1, which indicates an unused performance potential. The data sheet shown was generated during data collection at industrial companies and thus represents real manufacturing scenarios. In the investigated workpiece, this revealed potentials that reduce the machining time by around 13%. Applied to the company’s entire product portfolio, this would result in annual savings of around 18 MWh and 8 t CO2 per machine tool, assuming 3-shift operation on 6 working days, an average machine power consumption of 18.8 kW and a CO2 emission factor of 438 g/kWh (in 2019) [4, 5].

The tests proved a correlation between the real cutting torque and the current consumption of the spindle drive called up internally in the machine. However, these depend on the process parameters as well as on the spindle type and parameterization, which complicates the comparison between different machine tools.

3 Calibration Methods for Spindle Drives

In order to realize the comparability between processes and machine types, a transformation from a parameter- to a load-controlled process evaluation has to be carried out. This is to be achieved by means of a calibration procedure in which the characteristics of various spindle designs are recorded and a calculation methodology for the machine-internal data is created. In view of the fact that the calibration procedure has to be carried out on every existing machine tool and that acceptance in the industry must be ensured for this purpose, the implementation should not be associated with a high expenditure of time and machine downtimes. This enables the transfer of machine-specific data to an absolute reference system, which allows comparability and thus transferability.

For the necessary calibration of the current consumption of the machine internally derived spindle in the manufacturing production environment, various methods were investigated and checked for their suitability. Three approaches have been defined for the basic mode of action of the process:

Measurements:

In this case, the torsional moment generated in the machining process is measured directly and finally compared with the current consumption formed out of machine-internal data. This approach includes among others strain gauges, which measure the deformations resulting from the process forces and are either used in sensory tool holders or integrated directly into the corresponding tool spindle [6]. Adjustment: In these processes, a clearly defined torque is applied to the tool spindle. This can be realized by calibrated eddy current, hysteresis brake or agitators. Eddy current brakes are used, for example, on engine test benches to apply a defined load to the units [7]. Experimental series: The third category consists of defined metal cutting processes in which the machining forces and torques that occur have been determined beforehand. These test processes are implemented on the respective machine tool and the resulting recorded spindle currents are then compared with the target values.

Fig. 2.
figure 2

Suitability of the criteria for calibration methods, Legend: 1 = low/bad; 2 = medium; 3 = high/good

Full size image

Different requirements are placed on the corresponding calibration procedure with regard to application possibilities and process influence. On this basis, 8 criteria were developed and evaluated according to the approaches in Fig. 2, among others: Rotational speed: Scalability in terms of speed; Process impact: Influence on the real machining process; Application options: Applicability to different machining processes.

However, since the various criteria established cannot be regarded as equivalent in terms of weighting, they must first be ranked by means of a hierarchy. This was realized with a 3-stage pairwise comparison through which the corresponding weighting factors were created. These were used to offset the respective values in Fig. 2. This was calculated with the following formula:

$$ {\text{Rating}}\,{\text{ Criteria }}\,* \, \,{\text{Weighting}}\,{\text{ factor}}\, \, = \,{\text{ Evaluation }}\,{\text{number}} $$

The resulting overall evaluation including the weighting factors is shown in Table 1 and thus evaluates the suitability of the respective approaches “Experimental series”, “Measurements” and “Adjustment” explained before.

Table 1. Evaluation including weighting factor
Full size table

The overall evaluation shows that, with regard to the selected criteria, the “Experimental series” with 2.57 out of a maximum of 3.0 evaluation points is the most suitable for representing practical machining processes and for ensuring the necessary transferability to other machine types and configurations.

4 User Requirements

In addition to the technical challenges of collecting and processing machine data, such as a multitude of controls and machines, and the necessary measured value qualities, there are also the human requirements. In particular, the extensive implicit process knowledge of experienced and well-trained machine operators can only be used in a harsh production environment if an intuitive and efficient communication option between machine and user is realized. This must provide feedback on potential savings, stable process sections and prediction of change effects. At the same time, communication among all actors requires the consideration of systemic factors. This requires the collection of objective process and measurement data as well as information on selected optimization approaches and boundary conditions.

For all applications of optimization tools at the shop floor level, it is therefore crucial that the entire user experience is consistently practice-oriented. Only if this is ensured will an assistance system be sufficiently accepted by employees on the shop floor, even in areas with high performance pressure [8]. Requirements analyses carried out in manufacturing companies to date have identified one application area as particularly serious, from which specific user requirements arise.

Especially for productions of small and medium batch sizes, no elaborate studies for the optimal setting of a process are realized. The decisive factor is often the experience of the skilled machine operator. Manufacturing processes are usually optimized only as far as necessary to be able to produce economically. Here, a data basis is missing to make objective statements about the tool utilization in a process. A target group-oriented visualization of existing tool loads already opens up significant optimization potentials within a process.

In addition, due to the prevailing time and performance pressure, solutions and results achieved are rarely documented in such a way that other employees can apply this expertise to similar problems. Therefore, documentation close to the machine, at best fully automatic, must be able to record process parameters and malfunctions. This secured expertise increases productivity and permanently secures the company the corresponding process knowledge of the executing employees.

These requirements can be met with an optimization system that provides in-machine sensor data, enabling objective process evaluations, while also having an effective human-machine interface that allows employees to combine their expertise and the available measurement data. In this way, production processes can be optimized efficiently, taking into account all necessary factors.

5 UX-Example

The following graphical user interface (see Fig. 3) shows the optimization of the productive time of a milling process as a use case with the largest identified savings potentials. Human-centered design takes place through the four phases of identifying the user context, specifying user requirements, creating design solutions and evaluating the design solutions [9].

The increase in efficiency is achieved by utilizing the tools as evenly as possible throughout the entire production process. Necessary load changes are realized by local feed adjustments. Research and feedback from machine operators has shown that the particular information required highly depends on the process and product. In this way, the irrelevant data is hidden in order to provide the machine operator only with the essential information for process optimization. For example, the spindle speed is an irrelevant parameter in milling processes and is not shown in the dashboard in this case. The exemplary calculation of time, energy and CO2 savings, on the other hand, has a motivating effect and is therefore presented more intensively.

An optimization tool is available for the interaction, which narrows down the optimization range by defining a start and end point. The machine operator uses his extensive knowledge to include existing process restrictions in the optimization.

A visualization of the previously reached safe spindle load maximum serves as a reference line. Using drag-and-drop touch gestures, the programmed feed rate is adjusted locally by the operator. The expected effects on the spindle load are calculated using an approximate calculation according to Kienzle. By adjusting the color-coding, the operator receives direct feedback on the optimization potential. The granularity of the process section adjustment depends on the time available and the cost-benefit ratio of the process.

Through the user’s intervention in the pre-programmed process and the real-time prediction of its adaptation by the application algorithm, the user and the algorithm learn from each other in equal measure. Through feedback, users learn to refine their suggestions. The collected data can be used in the long term for partial automation, which enables a transfer from an experience- and parameter-controlled to a load-controlled process design and thus permanently optimal operating conditions.

Fig. 3.
figure 3

Assistant shows need for optimization (left), user defines action frame (top right) and performs spindle load adjustment by means of feed rate adjustment, impact prediction takes place immediately (bottom right)

Full size image

6 Summary and Outlook

These ongoing investigations carried out have shown that machine-internal data are suitable for process evaluation. To be able to transfer processes to other machine tools and thus make production much more flexible, the corresponding machine-internal data must be transferred to an absolute reference system by means of calibration procedures. The experimental series to be developed for this purpose must represent different processing scenarios and ensure the scalability of different parameters.Process monitoring and optimization based on machine-internal data offers an opportunity for metal-cutting processes to increase efficiency by utilizing existing performance reserves.

The appropriate visualization of this data makes it possible to use it effectively for process optimization by making optimal use of the machine operator’s knowledge and presenting the required information as comprehensibly as possible. Care must be taken to ensure that the visualization can be flexibly adapted to the respective target group to ensure that the crucial information is conveyed in a comprehensible manner.

In prospect, the various approaches are aimed at a wide range of stakeholders, from the requirements for machine-internal data and their interpretation to visualization for the user. This end-to-end approach creates a holistic view of the value chain of manufacturing processes. This is also shown by initial investigations in additive direct energy deposition processes and hand-guided welding processes. In this process, the various relevant process data, such as feed rate of the application nozzle or acceleration of the manual welding gun are examined and analyzed using an AI application. Here, targeted process monitoring can possibly also reduce the necessary quality assurance by means of costly and destructive testing and further increase process reliability.