Evaluation of in-line monitoring of micro-forming tribo-systems by consecutive U-bending

Robustness in micro-forming, a critical part of which is system monitoring, is important due to increasing demand for micro-components. In this work, a simulation of a micro-deep drawing tribo-system is applied to investigate the influence of die radius and production speed on process stability and tool wear. Monitoring of force and thermoelectric current are used as tools for wear detection, showing that both signals respond well to wear. The thermoelectric current is more sensitive to slight changes in process conditions and can be recommended as a wear monitoring method in micro-forming over using the force signal. The results of signal monitoring are compared to the surface roughness of formed parts as function of number of strokes, showing a good correlation.


Introduction
Demand for micro-components is increasing, leading to the necessity of adapting forming methods to their production [1].The use of existing knowledge and experience of macro-scale forming is not suitable for the design of micro-forming processes due to scale effects [2].The creation of new knowledge, specific to micro-forming processes, is therefore important to the continued development of robust production processes within micro-forming.
Tribology, the study of contacting surfaces in relative motion, is an important factor in any manufacturing process.A well-functioning tribosystem makes the difference between a stable process, and an unstable one.It is therefore of interest to ensure that tribo-systems are built in a way that allows for stable processes.This is especially important in the field of metal forming due to the number of components that are produced, and due to the combination of high contact pressure and relative sliding between tool and workpiece.Investigations of industrial tribosystems typically include some form of tribological testing, which allows for insights into tribological effects and process stability.Tribological tests are typically designed to be a simple representation of the industrial counterpart, reducing costs of testing, and allowing for more control over specific variables.Bay et al. [3] defined two types of commonly used tribo-tests in sheet metal forming, process tests and simulative tests.Process tests are inherently a better representation of the industrial system as they include the real process kinematics but allow for less control over specific variables and imposes a higher cost for testing.Simulative tests are a further abstraction of the process, leading to a weaker correlation to the forming process but a stronger control over specific variables and a smaller cost of running the tests.A U-bending test, in some works referred to as a strip-drawing test [4], is a direct process test in which a strip is drawn into a U-shape between two die-blocks.The test simulates the conditions occurring over the die radius in deep drawing [5], but neglects tangential compression, thereby simplifying the stress state in the workpiece.It has been used for continuous-stroke wear testing [6], testing of textured surfaces in hot stamping [7], and in studies on the influence of scale effects on springback [8] and is therefore an established test configuration.
Process monitoring in micro-forming inherently must be in-line due to the high production rate [9] as an out-of-control tribo-system can result in thousands of defective parts before production is stopped.Inline wear monitoring can be realized indirectly through monitoring of the quality of formed parts, or directly by monitoring of the tools.Direct methods, such as inspection of the tool surface by microscopy methods, are not suitable in micro-forming because of the small dimensions of the tools and the high production rate that the tools are operated at [10].Indirect methods are therefore preferrable, even though they give less information about the actual state of the tool.Monitoring of the force signal is commonly used, as changes in the force signal correspond to changes in frictional conditions.However, this is not always reliable in micro-forming as shown by Hu et al. [11], who found that a 100 % scrap rate due to tool-wear-induced defects in the formed part did not give a detectable change in the force signal.The use of thermoelectricity to monitor wear in-line in metal forming was pioneered by Wu et al. [12], who showed that a change in the thermoelectric signal corresponded to an increase in workpiece roughness.The increased roughness was then correlated with tool wear through optical inspection.They went on to show that when the thermoelectric circuit is connected through two tools, the sign of the thermoelectric signal indicates which tool is wearing [13].Other groups have investigated how the thermoelectric current affects processes such as blanking [14,15] and deep drawing [16], showing that the control of the thermoelectric current flow can affect the tendency for adhesive wear.
In this work, the monitoring of thermoelectric current and force development is used as a tool to detect wear in a micro-scale U-bending test.To the authors' knowledge, this is the first work applying thermoelectric wear monitoring in micro-forming.Wear is characterized by inspection of the tools and of the surface of the formed strip.As shown by Hansson and Liljengren [6], the occurrence of scratches on the drawn side of the strip can be taken as an indicator of wear, with heavy scratching preceding tribo-system failure.Two lubrication conditions, three die radii, and three production rates are applied to investigate the influence of each parameter on the performance of the overall tribosystem.

Experimental methods and materials
A U-bending test configuration was used for tribological evaluation in this work.A U-bending test involves the drawing of a strip into a Ushape as shown in Fig. 1(a).The strip is drawn in between two rectangular die-blocks while being bent and straightened over a given die radius.The test is a simulation of a deep drawing process, in which the stress state is simplified by neglecting tangential compression.An illustration of the specific test configuration that was used is shown in Fig. 1(b), and images of the test equipment is shown in Fig. 2. The test configuration consists of a pair of die-blocks (1), a punch (2), and blankholder (3) that are used to form the strip (4).An Eddy-current sensor (5) and target are used to measure punch location at any given time.Electrodes ( 6) are connected at specific points to record the thermoelectric signal, and a force transducer (7) is used to measure the force applied by the punch.A freely moving disc (8) sits behind the force transducer, allowing it to be preloaded by a screw (9).A spring-loaded ejector pin (10) is used to clear the formed geometry from between the die-blocks between strokes.A Teflon sheet (11) is used to isolate the active tool area from the surrounding sub-press, reducing noise in the thermoelectric current signal.
Two forming machines were used to actuate the tests.All tests involving a production rate of 120 strokes per minute (spm) were performed on an EPK-32 eccentric press from PMB at the Technical University of Denmark.The press, which has a load capacity of 320 kN.The punch location was measured in the EPK-32 press using an HLP190/ SA linear potentiometer from Penny & Giles that was mounted to the press ram.The stiffness of the press was assumed to be high enough to allow for accurate measuring of the location of the punch from the location of the press ram as the forming load is small.Tests involving higher production speeds were performed on a BSTA 810 stamping press from Bruderer at the Institute of Production Engineering and Forming Machines at the Technical University of Darmstadt.An EU8 Eddy-current sensor was integrated into the U-bending tester and used to measure the punch location during these tests.
The basic configuration of the experimental setup used in both presses consists of several key systems and is illustrated in Fig. 3.The sheet metal strip is pulled by the feeder from right to left.The coil of strip sits on a de-coiler, with slack being ensured between the strip tensioner and the coil of strip.A small amount of tension is applied between the coil and U-bending tool by the strip tensioner to ensure consistent feed and consistent tension in the strip.The strip is then guided through a lubrication system which applies lubricant to the top and bottom of the strip.The oiled sheet metal strip is then fed through a strip guide and into the forming tool.After forming, the strip is led through another strip guide to ensure consistent positioning of the strip as it is formed, and the U-shaped geometry is then crushed to allow easier handling.The formed strip is then pulled by the feeder, and eventually re-coiled for storage.
The punch force and thermoelectric current were acquired during all testing.The force was measured by a Kistler 9101C force transducer that was mounted directly above the punch so that the entire forming force would be transferred to it.The force transducer was pre-loaded to a load of 5 kN to improve system stability, and it was calibrated before testing after installing it in the tool.The signal from the force transducer was amplified in a Kistler 5015 charge amplifier and acquired by an NI 9215 module at a sampling frequency of 2-5 kHz.The thermoelectric current was acquired through two low-resistance copper cables that were attached to the punch and die-holders, respectively.The cables were connected to an NI 9208 module that was used to acquire the signal at a sampling frequency of 500 Hz.
The active part of the stroke, controlled by the movement of the press-ram, was constant at 5 mm, with the strip being moved 30 mm between strokes.The blankholder force was applied by four springs that were pre-compressed by 5 mm.The springs have a spring constant of K = 22.9 N/mm, so the initial blankholder load was 458 N which had increased to 916 N at the end of a stroke.
The strip, which had a cross-section of 14 mm × 0.1 mm, was made from AISI304 stainless steel that had been tempered to half-hard condition.The flow-curve of the material, measured by plane-train stack compression of the raw strip [17] without accounting for friction, is shown in Fig. 4(a), and the surface of a fresh strip is shown in Fig. 4(b).
The fresh strip has a surface roughness of Sa = 0.257 μm ± 0.013 μm, Ra All tools were made from Vanadis 8, a tool-steel from Uddeholm [18], that had been hardened to 64 HRC.The tools were EDM-wire cut to the geometry shown in Fig. 5(a), and their surfaces left with the EDM texture shown in Fig. 5(b).The surface roughness of the tools was measured optically and found to be Sa = 0.201 μm ± 0.04 μm, or Ra = 0.154 μm ± 0.017 μm.Die radii of 0.15 mm, 0.30 mm, and 0.50 mm, having a variation in the radius on the order of 2 μm, were used in this work.The punch was 20 mm wide, so that the entire strip width was used in testing.Die-blocks have two rounded edges, but due to wear of the top surface only one of these was used on each die-block.The other radius ensures that lubricant is not wiped off as the fresh strip enters the tooling.
A lubricating emulsion, 1%v/v Rhenus FU60 mixed with 99%v/v tap water, was used as a lubricant.This lubricant is already used for industrial micro-deep drawing processes.It was applied by a felt roller system from Raziol.Two lubrication conditions were applied, one in which lubrication was supplied during the whole test, and another in which lubrication was only applied for the first 1800 strokes, after which it was stopped.These test conditions are referred to as LTx or DTx respectively from hereon, with x indicating the test number without a change in test parameters.In the latter condition, residual lubricant in the felt roller meant that the strip was lubricated with a decaying amount of lubricant until a total of around 3000 strokes had been performed.The lubricant amount used in the PMB press was enough to ensure abundant lubrication.In the Bruderer press, lubricant was applied using a spray-nozzle system from Raziol at a rate of 5 ml/min, which gave a coverage of approximately 10 g/m 2 on each side of the strip.

Signal monitoring 3.1.1. Individual strokes
Examples of measurements from select strokes at intermittent points throughout tests are shown in Figs. 6 and 7.The data were taken from a test in which an R0.3 die-block was used under the partially dry condition (DTx, x being the repetition number) at 120 spm for a total of more than 10,000 strokes.As shown in Fig. 6, the force decreased slightly between the 50th stroke and the 3000th stroke due to run in while the tribo-system was lubricated.This effect led to partial smoothing of the surface caused by pressure peaks that deform the tool material, adhesive transfer into surface valleys, or the formation of a boundary lubrication layer.The result was a more favorable tool surface that led to a reduction in friction.The force remained stable for a period, but then increased with further strokes due to wear.The thermoelectric current signal followed the same trend as the force signal.stroke 50 to stroke 3000 shows that the signal amplitude decreased slightly, indicating less heat generation in stroke 3000 compared to stroke 50.This is consistent with the discussion of the force signal, indicating that the thermoelectric signal could be used to detect run in.After the lubricant was depleted, i.e., after 3000 strokes, the amplitude of both profiles started to increase.While the force at initial contact (IC) was constant, the peak force and the slope from contact to bottom dead center (BDC) increased.The increase in the force profile was caused by the sliding friction restricting material flow increasingly due to wear.
The bottom of the thermoelectric current profile moved down during the sliding, i.e., before reaching the BDC, as a function of wear.This indicates that more heat was being generated at one end of the circuit, in this case in the die-blocks.It is difficult to comment on the thermoelectric signal during retraction of the punch, i.e., after the BDC, but the increase in the signal for the 6000th and 10,000th strokes compared to the 50th and 3000th strokes may indicate that the punch experienced increased friction during retraction.This is consistent with adhesive build-up on the tools, which reduced the size of the die opening.In further analysis, only the force peak and minimum thermoelectric current signal (|min(I th )|) per stroke are considered as these values represent the state of the system.

Lubrication condition
Fig. 8 shows the results of testing with both lubrication conditions for all die-blocks.Tests are color-coded from hereon, so that the colorscheme is consistent from Fig. 8 until Fig. 11.Consistently, no signs of wear could be detected in either signal for tests that were lubricated for the whole test duration of 10,000 strokes.Signs of the force and thermoelectric amplitude decreasing could however be seen, indicating running in of the tools.The force signal showed signs of occasional instability, but this was not observed in the thermoelectric signal and not investigated further.The thermoelectric amplitude was more stable over time.The results of tests that were run partially dry showed an increase in both signals over time due to wear of the tools.It should be noted that the test labelled DT2 for the R0.15 die radius experienced early fracture of the strip before the intended 10,000 strokes could be  performed.It is clear from the plots shown in Fig. 8 that there is a correlation between an increase in the force and an increase in the thermoelectric amplitude in the wear monitoring.

Die radius
Fig. 9 shows the effect of die radius on the wear development of the system.As shown in Fig. 8, the occurrence of wear is reproducible.To ensure clarity in plots, only results from one test of each test series is plotted here.From the start of a test, it is clear that a smaller die radius leads to larger drawing forces and larger thermoelectric amplitude, as more heat is generated in the die-blocks.With little or no wear, no significant changes were detected in the profiles for the lubricated tests.However, as soon as lubrication was stopped, the signal values grew.The force values grew at the same rate for all die radii, indicating that the wear development is similar.However, the thermoelectric signal grew fastest for the smallest die radii, and slowest for the larger die radii.The heat being generated per stoke is therefore increasing more rapidly for the smaller die radii, showing that effects that do not show up in the force signal are caught by the thermoelectric signal, so that the thermoelectric signal is more sensitive to wear development on the tool surfaces.Further, small radii wear faster than large radii, in that small die radii lead to less stable process kinematics as shown by the more rapid increase in the thermoelectric signal.No significant wear could be detected for the fully lubricated tests by considering either the force or thermoelectric signals, although the process severity increased for smaller die radii as it did for the partially dry tests.

Forming speed
Tests performed on the PMB press were performed with a total stroke-height of 20 mm, whereas a stroke-height of 16 mm was used in tests performed on the Bruderer press.The 120 spm, 300 spm, and 600 spm production rates thereby corresponded to average forming speeds of 80 mm/s, 160 mm/s and 320 mm/s, respectively, with the active forming stroke length being the same for all tests.Fig. 10 shows the results of tests performed at the three different production speeds, applying the partially dry condition.The force was the same while lubrication was being applied, showing that the force was not affected much by forming speed in the range applied in this work.However, the thermoelectric signal shows a large difference between the different tests.As the tests performed at higher forming speeds allowed for a shorter time for heat to be generated and a shorter time for heat to be conducted away from the tool interface, the temperature gradient in the tools was larger, and thereby the thermoelectric signal also.Lubrication was stopped at different points in these tests, with lubrication in the 120 spm test being depleted at approximately 3000 strokes, but later for the 300 spm and 600 spm tests, respectively.The choice of point for stopping lubrication was based on allowing the system enough time for running in and achieving steady-state.Without allowing the system to run in, the strip fractured immediately as soon as lubrication was stopped for tests applying higher production rates.
At lower forming speeds, the force and thermoelectric signal increased gradually.However, for the higher forming speeds, this increase was much more abrupt.As shown by the close-ups in Fig. 11, it took approximately 600 strokes for the strip to fracture in the 300 spm test, whereas it took approximately 200 strokes for fracture to occur in the 600 spm test.The increase in forming speed led to a higher steady state temperature being established in the system, so that the system was more sensitive to a lack of lubrication.Compared to the force profile, the increase in process severity is much clearer in the thermoelectric signal.This indicates that it is more sensitive to wear events than the force signal and might therefore be more useful for process monitoring.

Tool/workpiece inspection
As shown in Section 3.1, very little wear occurs in the fully lubricated tests while wear was detected in the partially dry tests.An example of a die edge before and after a test performed in the partially dry condition is shown in Fig. 12. Fig. 13 shows close-ups of the worn and unworn areas of the die-block edge.Clear visual differences can be seen between the light and homogenous unworn radius and the darker worn radius.Two wear mechanisms are at play on the die radius.Firstly, the geometry itself is worn causing the radius to grow.This should lead to a decrease in the force, as also shown by comparing the force measurements for different radii of the die edge.The second mechanism, adhesive build-up in asperity valleys on the edge of the die, led to an overall smoothing of the surface as the tool is run in.Eventually, as wear progresses, build-up continues and starts to change the shape of the surface, eventually leading to scratching of the formed part and fracture of the strip.The latter mechanism is clearly dominant in this case as the force grows as a function of increasing wear.Further characterization of the surface of worn tools did not show any change, e.g., in the surface roughness.This showed that the process more readily results in strip fracture for small changes to the tool surface, highlighting that changes in friction and other parameters in micro-forming processes are more critical to process stability than those in macro-scale processes, as also shown by Vollertsen et al. [19].Further investigation of the tool surfaces was not performed due to the small size of the tools, which also makes this difficult to do in practice.
Tool wear in a U-bending test can be detected by inspection of the formed part.An increase in the roughness of the surface perpendicular to the drawing direction is a clear indicator of wear, as scratches are formed [6].In this work, the test labelled DT2 in Fig. 8(a) was selected for inspection of the formed strip.Images across the width of the strip were acquired at certain stroke intervals using an Olympus LEXT laser confocal microscope.Selected, acquired images of the undeformed strip and the formed strip are shown in Fig. 14.There is a clear increase in the amplitude of the surface with an increasing number of strokes.Comparing (b) and (c) shows that the height slightly decreases, which is consistent with running in, and then the height increases in (d) and even further in (e).Comparing these surfaces to the fresh strip shows that the rolling texture is removed during drawing, potentially due to strain-induced surface roughening [20] which would occur due to increasing friction in the die-strip interface.
Nine surface profiles were extracted across the width of the strip from each image shown in Fig. 14   shows that the surface roughness decreases slightly in the beginning, as might be expected during running-in, although the variation in the surface roughness shows that this may be lost in the noise of the measurements.Later, as the tool surface wears, the surface roughness increases along with the force.The increase in force is due to wear of the die, although changes to the die surface were not clear when the dies were investigated.The development of the surface roughness follows the development of the force and thermoelectric signal, although it took much more effort to determine it.The use of intermittent evaluation of surface roughness of formed parts is therefore a method that can be used  to monitor wear, as also shown by Hansen and Liljengren [6].However, due to the low time resolution of this method, it should not be used unless no other options are available.

Conclusions
In this work, different methods of wear monitoring were evaluated by application in a U-bending test configuration that was designed for use with thin foil workpieces.Force, thermoelectric current, and measurement of surface roughness were compared for different die radii, different production speeds, and different lubrication conditions.Based on the discussion and results presented in this work, the following conclusions were drawn: -The use of thermoelectric signal monitoring force monitoring is suitable detection of wear in micro-forming processes.-Die radius has a large influence on process severity in terms of friction and wear.Smaller die radii lead to higher contact pressure, and thereby higher friction and drawing force.This was clear based on the difference in force and thermoelectric signal for different die radii.
-Forming speed has a small influence on the process force, showing that it does not affect friction and wear much at speeds tested here.However, the increase in the thermoelectric signal shows that the heat generated on a per-stroke basis is higher.This will lead to a higher steady-state temperature, and thereby is likely to cause additional wear.-The thermoelectric signal is more sensitive to changes in process conditions, such as lubrication stopping or die radius changing, than the force signal.-The development of wear according to the force and thermoelectric signals followed a similar behavior as the development of the surface roughness of the formed strip.-The use of inspection of formed parts is a valid approach to monitor processes.However, it is not suitable for frequent inspection of micro-forming processes due to the necessary time for inspection and the small amount of information that results.

Fig. 1 .
Fig. 1.U-bending tester shown by (a) the principle of the test and (b) an illustration of the equipment used in this work.

Fig. 2 .
Fig. 2. Test equipment used in this work.Units are in mm.

Fig. 5 .
Fig. 5. Die-blocks used in U-bending shown by (a) base geometry and (b) a typical tool surface with wire-EDM texture.

Fig. 6 .
Fig. 6. Results of partially dry test (DT1) using an R0.3 die block shown by force as function of time for select strokes.Error bars indicate standard deviations of peak force for ±5 strokes from nominal stroke number.

Fig. 7 .
Fig. 7. Results of partially dry test (DT1) using an R0.3 die block shown by thermoelectric current as function of time for select strokes.Error bars indicate standard deviations of minimum thermoelectric current for ±5 strokes from nominal stroke number.
to evaluate the surface roughness in terms of the Ra and Rz parameters.The surface roughness across the strip, perpendicular to the drawing direction, as function of number of strokes is shown in the bottom of Fig. 15.The peak force and thermoelectric signal per stroke were plotted to facilitate comparison and are shown in the top of Fig. 15.Comparing the top and bottom of Fig. 15

Fig. 8 .
Fig. 8. Results of lubricant condition testing for different die radii shown by (a,c,e) peak force per stroke and (b,d,f) thermoelectric amplitude per stroke.Dashed line indicates when lubrication was stopped for DT tests.

Fig. 9 .
Fig. 9. Comparison of results from different die radii.(a,c) Peak force from dry and lubricated tests, respectively, (b,d) thermoelectric amplitude from dry and lubricated tests, respectively.Dashed lines indicate when lubrication was stopped for DT tests.

Fig.
Fig. Results from tests performed at different production speeds shown by (a) peak force per stroke, and thermoelectric amplitude as function of number of strokes.Dashed lines indicate when lubrication was stopped for measured data from tests shown in the same color.

Fig. 12 .
Fig. 12. Example of (a) fresh tool edge and (b) worn tool edge.Note that scale bars are accurate along the tool edge, but less accurate across it due to the curvature of the tool edge.

Fig. 13 .
Fig. 13. of die-block used in test labelled DT2 in Fig. 8(a,b) shown by (a) unworn edge, (b) end of worn edge area, and (c) middle of worn edge area.Note that scale bars are accurate along tool edge, but less accurate across it due to the curvature of the tool edge.

Fig. 14 .
Fig. 14.Surface of (a) fresh strip and strip formed in test labelled DT2 in Fig. 7(a,b) after (b) 100 strokes, (c) 2500 strokes, (d) 4900 strokes, and (e) 6600 strokes.It should be noted that only a 6 mm × 0.6 mm region of the center of the strip is shown.

Fig. 15 .
Fig. 15.(top) peak force and thermoelectric signal compared to (bottom) surface roughness of formed strip as function of number of strokes.