Towards Real Time Predictive System for Mechanical Stamping Presses

to Assure Correct Slide Parallelism

Ivan Peinado-Asensi

1,2 a

, N. Montes

1 b

and E. Garc

ıa

2 c

Mathematics, Physics and Technological Sciences Department, University CEU Cardenal Herrera,

C/ San Bartolom

e 55, 46115, Alfara del Patriarca, Valencia, Spain

Ford Spain, Pol

ıgono Industrial Ford S/N, 46440, Almussafes, Valencia, Spain

Keywords:

Smart Stamping Plant, Real-time Data Analysis, Machine Health Monitoring, Predictive Maintenance.

Abstract:

Automotive companies are going through a rough time due to the decrease in the car sales market, therefore

OEMs trend is cost reduction in the next years over improving efﬁciency increasing digitalization, implement-

ing new industry 4.0 technologies to turn their facilities in smart factories. Within car manufacturing processes,

stamping present many possibilities for development, in this paper an approach to bring stamping plants closer

to smart factories is presented. The most common problems in stamping are unexpected breakdowns in equip-

ment and poor quality parts produced, to avoid these problems corrective and predictive maintenance tasks are

carried out to improve presses and tools performance. One of the critical maintenance tasks in press machines

are parallelism, a malfunction in the kinetic transmission can lead to high cost and duration breakdowns. To

monitor machine working parameters a novel method is presented using IIoT techniques, having access to

machine working parameters in Real-Time to predict machine malfunction in order to reduce the number of

breakdowns.

1 INTRODUCTION

Monitoring and Controlling press processes in the in-

dustrial stamping world has been widely developed to

ensure machinery lifetime and high quality in the ﬁ-

nal product. But we are still having daily breakdowns

in the production lines in the factory due to equip-

ment failure and defects in the ﬁnal product manu-

factured such as splits, necking and wrinkling among

others. The previous research in control processes has

followed four possible paths (Lim et al., 2013) to im-

prove performance. One of them is the die try-out

process where the objective is to analyze the tool (die

design) and the working variables. The second one

is controlling the blank holder force using ﬁnite el-

ements analysis (FEM). Other is based in in-process

control, a strategy to monitor the process inputs and

outputs during the stamping process. The last one is a

post-process control which part inspection is made to

identify signiﬁcant variables.

Stamping machines need periodical maintenance

tasks to ensure correct functionality and avoid the

https://orcid.org/0000-0002-5603-5910

https://orcid.org/0000-0002-0661-3479

https://orcid.org/0000-0002-4210-9835

mentioned breakdowns, we need to know the com-

ponents status and ﬁx them if required. Parallelism is

one of the main maintenance tasks that require time

and resources to carry it out and it is also important

to ensure a correct slide motion. The imbalance in the

slide is detected by the parallelism maintenance task,

this is carried out by setting the press in the Bottom

Dead Center without the die, placing the four posi-

tion gauge under each slide corner and next raising

the slide counterbalance pressure to lift the slide up

and then drop it to the initial position again, this is the

process carried out to measure slide clearance. Once

the clearance is known, again in the BDC we take a

slide corner as a reference and measure the distance

of every corner to the ﬂoor, thus it is possible to know

the balance difference in the corners. Once it is mea-

sured, the connecting rod feet are adjusted until the

parallelism of the slide is balanced. As explained, the

procedure is complex and requires a lot of time and

resources.

In this research we are going to make an approach

applying an in-process control methodology exploit-

ing the amount of data available through connected

devices employing Industry 4.0 technologies (e.g., In-

ternet of Things, Big Data Analysis) monitoring the

396

Peinado-Asensi, I., Montes, N. and García, E.

Towards Real Time Predictive System for Mechanical Stamping Presses to Assure Correct Slide Parallelism.

DOI: 10.5220/0010547203960402

In Proceedings of the 18th International Conference on Informatics in Control, Automation and Robotics (ICINCO 2021), pages 396-402

ISBN: 978-989-758-522-7

variables of the machine, where exist up to 40 dif-

ferent variables to measure, among all the available

data we will procedure to identify and classify the

variables that are considered crucial in the process,

directly related with the slide parallelism to obtain in-

formation to identify working pathologies. In the fol-

lowing sections it will be explained which variables

and why are gonna be used to solve the parallelism

problem, with the goal of having a full knowledge

of the working state of the press in real-time. The

problem statement is to verify if it is possible to de-

tect the slide imbalance without requiring parallelism

maintenance task, skipping the process and the use of

the tools and sensors that are required to make it as

the procedure presented in (Magraner, 2016). To do

that a real-time monitoring tool has been developed

to know the slide gravity center and the friction in the

gibs. If the friction increase in some of the gibs or

the slide centre of gravity is displaced we could en-

sure that the parallelism is not correct. Once the vari-

ables are selected among the forty available, will be

of vital importance to verify if we can achieve to do

accurate predictions. Afterwards it will be needed to

deﬁne stamping working patterns to verify if the mon-

itored variables are reporting appropriate information

of the parallelism, if not it will requires to evaluate

new ones.

The purpose of the research is to know the slide

Gravity Center (GC) in order to control parallelism

and avoid component wear due to malfunction with

a predictive tool fed with data measures in Real-

Time following a procedure called Maintenance Sup-

port System (MSS) (Garc

ıa et al., 2018), and once

achieved, the following procedure will be to continue

monitoring new variables that cannot be measured us-

ing the usual procedures. In this paper a predictive

tool is proposed to know the GC of the slide in a cy-

cle motion of the press using the signal data from the

strain gauge available in the machine without required

installation to monitor the process in real time. The

motivation is to know the press working performance

through the GC in every moment, being able to know

the quality of the stamp punch, overloads, deviation

of the slide parallelism and the slide travel motion to

avoid premature wear due to friction in the press com-

ponents.

The paper is structured as follows: In section 2, an

overview of the research done in control and monitor-

ing in industrial stamping. In the next section 3, we

will explain the proposed physical model proposed

with the data required to get the friction variable. In

section 4 there is an explanation of how the data is

monitored and the result of the models using real-time

acquired data. In section 5 we will show the devel-

oped tool to ﬁnd out the parallelism in real time. Fi-

nally, in section 6 we show the future work and the

conclusion.

2 MONITORING STAMPING

PROCESS

2.1 Equipment Description

Two main types of presses are used in sheet metal

forming, mechanical (Single-action (SA) or Double-

action (DA)) and hydraulic, both can be used in

all operations carried out in stamping (blanking,

deep-drawings and trimming) (Wagener, 1997). In

Ford Spain stamping plant most of the presses are

mechanical and the experiment of the research is

being carried out in a mechanical SA press with a

cushion system. The major difference between SA

and DA presses is the eccentric drive transmission

system and the blank holder force system. In SA

presses as seen in ﬁgure 1 there is one slide for the

tool and a cushion system as blank holder unlike DA

presses that have an eccentric drive system with two

slide displacements, one for the blank holder and

other for the tool holder. Double effect presses have

been used during years for deep drawing operations,

but recently they have been replaced by hydraulic

and SA presses with cushion because they are more

effective in deep-drawing operations (Hoffman,

1998).

As seen in ﬁgure 1, a stamping press is manufactured

of many different components such as the electrical

engine which activates press motion, the eccentric

transmission system which moves the slide upward

and downward controlling the speed press, the clutch

and brake hydraulic system, the slide counterbalance

cylinders that help electrical engine to raise the slide

and keep the torque of the activation system balanced

in the cycle, the slide where the die is ﬁtted for each

production and height can be adjusted and also the

gibs in each of the press columns to guide the slide.

2.2 Process Control in Sheet-metal

Forming

There is much research based in methods that they

use data for monitoring working process, this research

mostly in close-loop and in-process control systems.

In last decades it has been easy to ﬁnd research about

close-loop control as the one developed by (Siegert

et al., 1997) (Yagami et al., 2004) (Viswanathan et al.,

Towards Real Time Predictive System for Mechanical Stamping Presses to Assure Correct Slide Parallelism

397

Figure 1: Single-action mechanical press.

2003) in which the springback effect of the material

is controlled by the Blank Holder Force applied, con-

trolling the material ﬂow when deep-drawing (Endelt

et al., 2013) and also eliminating the wrinkling and

fracture in rectangular parts (Ahmetoglu et al., 1995).

Also non-destructive test (NDT) have been used to

monitor data, in the case of (Ng et al., 2007) to di-

agnose the stamping quality, the 3D thermal distribu-

tion was analyzed, using the electromagnetism physi-

cal principle by installing electromagnetic coils in the

die surface (Shang and Daehn, 2011). There is much

research based too on Acoustic Emission in order to

detect bad parts (Song et al., 2016) or to detect tool

wear (Ubhayaratne et al., 2015).

Regarding to in-process control systems, in the

study carried out by Hardt and Fenn it was possible to

detect differences in material input such as lubrication

and material thickness (Hardt and Fenn, 1993), that

was a breakthrough in control monitoring because it

was possible to identify new variables in stamping

process. Hardt et all. used the data monitored from

the force punch combined with a previous close-loop

control developed to control the blank holder force.

Other effective process control designed was the one

proposed by (Hsu et al., 2002) in which they take into

account blank holder pressure, the sheet metal fea-

tures and the punch force showing that the process

can maintain the same punch force trajectories with

different lubrication conditions, but machine control

cannot.

Figure 2: Process control of sheet metal forming.

The proposed procedure is focusing on taking a

step forward, the idea is following in-process tech-

niques by monitoring the required data to analyze

working performance using Industry 4.0 technolo-

gies, but to carry it out a huge investment is needed

in order to buy and install proper equipment to get ac-

curate data and centralize the measurement system in

one platform. Furthermore, due to the currently situ-

ation in the automotive sector the OEMs are planning

to reduce costs. It is true that by increasing digitaliza-

tion and optimizing the efﬁciency of productions the

costs can be reduced, but a large initial investment is

required anyway. Therefore, we are going to apply

Industry 4.0 techniques avoiding this initial invest-

ment by acquiring the data from the sensors already

installed in the presses of the plant, and by develop-

ing the software in site to analyze the data and send

alarms to predict breakdowns.

3 MATHEMATICAL MODEL TO

MONITOR STAMPING

PROCESS DATA

Taking as starting point that the bad parallelism and

the imbalance in the transmission system leads to a

premature wear of the different components in the

press and this wear is caused because of the friction

between them, we can conclude that friction and im-

balance are directly relate. The aim of the data ac-

quisition is to obtain curves of all the available mea-

sured variables in the press, at least on value per each

press position, this means 360 values in a curve shape

of tonnage force, counterbalance pressure, overload

pressure, cushion pressure, etc. Therefore relating all

this data in a physical model could be possible to ob-

tain the friction force of different components.

Thus a physical model will be proposed in this

section to monitor the friction force variables in a

press. The model is going to be set in three stages.

Firstly, a model describing the downward motion

where the friction in the gibs and in the eccentric

transmissions systems can be obtained. Secondly, the

deep-drawing where the friction of the die surface

with the blank surface could be modelled and in the

ICINCO 2021 - 18th International Conference on Informatics in Control, Automation and Robotics

398

third stage, the upward motion of the slide where the

same friction force can be obtained as in downward

motion. So following the procedure used to monitor

the GC and thanks to the data acquisition system, a

physical model is presented to know the friction in the

slide gibs in real-time during the downward motion.

In picture 3 the loads considered in the model are

shown. The proposed model will be used to ﬁnd

out the friction force in each press corner. So if we

consider the system as static in every press position

degree, where there is no downward movement and

therefore we will consider there is no friction in the

gibs, so for one corner it results,

− F

· a

, i = 1. (1)

Where F

is the slide counterbalance force, F

the

press force applied in the slide measured with the ton-

nage gauge sensor, m the upper die mass and a

the

acceleration of the slide in the downwards motion.

The model is used to describe the physical behaviour

in every corner of the press as described in picture 3.

Figure 3: Slide loads considered.

− F

· a

, i = 1. (2)

Considering the forces applied by the components in-

volved in the downward motion of the slide being F

the gib friction force. Now it is shown the result of

the physical model for the entire mobile system,

∑

i=1

− F

) +

∑

i=1

= m · a

(3)

As it can be seen, the only parameter that is not known

is the friction force, which will be obtained from the

model proposed. Applying this equation to every

press position degree from 320

to 90

we are able

to control the friction force existing between the slide

and the gibs.

4 DATA ACQUISITION PROCESS

The standard which has already been implemented by

Ford in Valencia Factory to program the PLC to get

the data explained in (Garcia and Montes, 2019) al-

lowed to get direct connection to the already installed

sensors by the press manufacturer. Saving money in

the installation because the sensors to get the info are

already installed in the press except the acceleration,

the only variable needed that was not previously mon-

itored but whose installation was really easy as ex-

plained below.

The challenge at this point is to have enough scan

speed in the PLC to read all the required data in a

stamp cycle, so due to the slide motion it is carried out

in more or less 3.5 seconds and 360 values per sensor

which need to be read by the PLC and if there are nine

variables it is exactly 3240 values, it requires around

1 ms second of scan speed per cycle. After several

tests carried out in the installed PLC, it was proved

that no data was missing in the cycle. Therefore, the

data collection started to obtain the curves of the data

per cycle to calculate the friction force.

To know the slide acceleration parameter, of vital

importance in the proposed model, a cheap and sim-

ply solution has been carried out, having access to an

important variable without signiﬁcant expense. An

accelerator sensors has been installed (detail in ﬁgure

5) in each corner of the slide, having the acceleration

in the whole motion cycle, the sensor was connected

through a switch to the PLC to be able to collect all

the data where the other variables are monitored. The

tonnage data is acquired from the gauge sensors in-

stalled in the columns of the press, there are four

gauge sensors installed which send the data through

the Helm switch to the PLC. The slide counterbalance

force has been read from the pressure sensor installed

in the pneumatic circuit of the system. There is just

one measure that is why in the model is divided by

four, taking it as proportional in each corner.

5 REAL-TIME PREDICTIVE

TOOL

The online tool presented is currently working with

tonnage force data, the press is equipped with a load

gauge sensor in each one of the connecting rods which

allow us to know the press force applied in the cy-

Towards Real Time Predictive System for Mechanical Stamping Presses to Assure Correct Slide Parallelism

399

Figure 4: Gravity Center in Real-Time.

cle. As far as it is known with the force carried out

by the press in each of the four rods it is possible to

obtain the gravity center of the slide. By getting the

data from the PLC and applying the calculus in our

tool we can ﬁnd out the gravity center at every point.

The tool shows the gravity center in real-time at Bot-

tom Dead Center press position (180 degrees) where

the maximum press force is applied, depending on the

produced part on the machine where the experiment is

taking place the maximum force can be up to 2.1 ·10

(kg), close to the press limit which is 2.5 · 10

(kg).

Knowing the gravity center of a squared object where

some forces are applied is trivial, thus the process of

implementing it in the software, developed in site, and

also having the Ford standard platform to connect the

machine data to the data based in the cloud was not a

big deal, the result obtained is shown in picture 4.

Different areas are shown in the online platform 4,

at the top left you can see the GC displacement of the

slide in 2D at BDC press position, the red square in-

dicates the press working limit for that applied force

and the blue square delimits the size of the slide. At

the top right we can see the GC plotted in 3D, where

the z axis indicates the press force for each stamped

part. The graph at the bottom shows the data obtained

from the sensor signal selected. There is also a menu

where we can choose the date of the production when

the data needs to be analyzed, the sensor type needs

to be analyze and also to classify the data depending

on the manufactured body part, because depending on

the body parts produced the data can vary due to ma-

Figure 5: Accelerometer sensor location.

terial and die characteristics. There is also the option

to visualize the GC of a cycle curve as you can see in

picture 6.

The tool purpose is monitoring the gravity center

and check if there is any variation or pathology that

appears just in time which can give information about

the machine status. It is well known that a deviation

of the slide balance can lead to premature wear in the

eccentric motion transmission system, wear of the die

surface and also in the slide gibs, caused by friction

ICINCO 2021 - 18th International Conference on Informatics in Control, Automation and Robotics

400

Figure 6: GC of a cycle.

increasing that may cause a lot of problems in the pro-

duced parts and the machine function. Knowing the

GC information in Real-Time can give us an advan-

tage to reduce maintenance activity, reduce mechan-

ical wear and reduce the set up press working force

and avoid parallelism checking extra works, all of that

can be achieved directly. Indirectly, electrical con-

sumption and repairing time can be reduced, break-

downs can be predicted and the life of the equipment

extended. All the improvements listed can bring cost

reductions.

5.1 Health Machine Monitoring

The next step is to know when the machine stops

operating under normal conditions and to identify

the problem that makes this happen. To do that, we

will send an alarm when the GC deviates from the

predeﬁned boundaries. As seen in picture 7

Figure 7: X-Y section of limit surface force in 2D.

while the GC is still inside the green area we are

going to consider it to a normal function, once it

reaches the yellow area it will send an alarm to know

the behaviour of the press and check whether it is a

random hit with imbalance or a cumulative event of

hits out of the normal working area, if this happens

a technician will be sent to check the machine status.

When the hit reaches the red area the same procedure

as in the yellow area will be carried out but with an

exhaustive inspection and ﬁnally if it goes beyond of

the force limit line, the press will stop automatically.

0,00

500,00

1000,00

1500,00

2000,00

2500,00

3000,00

-3000 -2000 -1000 0 1000 2000 3000

Slide length

LIMIT FORCE X AXIS

Fx max Critical limit Cheking limit

Figure 8: X-Z section deﬁning status limits in X axis.

Here in picture 7 we deﬁne the limits to identify

when the parallelism is deviating, these limits are de-

ﬁned inside the red square at the top left graph in pic-

ture 4. To show the deﬁned alarm area more precisely

in picture 8 you can see the surface in 2D from a right

proﬁle view.

The same is deﬁned in y axis and the result of the

shape of the different alarm area volumes can be con-

sidered as proportional surface with reduced volumes

inside the force limit surface that you can see at the

top right in picture 4. From the 360 points of the GC

obtained per cycle it will considered the percentage of

the points that there are in different area. The result

of the example curve showed before in the picture 6

there is a 100% of the values inside the green volume.

6 CONCLUSION

The research will continue monitoring the proposed

data in the physical model that is left to measure, fol-

lowing the same procedure carried out on the tonnage

force (such as counterbalance force and acceleration)

in the whole cycle and simulate the stamping cycle

to get the friction force. The cycle is divided into

3 stages. The ﬁrst one, downward motion to know

the friction in the gibs, the second one is the same

in upward motion and the third one will be the deep-

drawing when more variables will be considered, such

as the the hydraulic cushion force, to get the friction

force between the die surface and the blank material.

To do that ﬁrst it is required to extract a big amount of

data, check if the PLC is able to sent the data without

Towards Real Time Predictive System for Mechanical Stamping Presses to Assure Correct Slide Parallelism

401

loss of information, for later process the information

in our predictive tool.

Using industry 4.0 technology a tool developed to

control the machine in real time has been created, fol-

lowing the OEM trend of transform their current man-

ufacturing facilities in smart factories. The tool devel-

oped works as expected in real-time and give accurate

and valuable information about the slide state work-

ing, been able to control parallelism and friction in

the gibs reducing maintenance activity and premature

wear altogether with reduction of electrical consump-

tion. Furthermore the developed tool result as a ro-

bust and powerful method that gives a lot of opportu-

nities when talking about predictive maintenance and

knowing the machine health, taking the lead predict-

ing breakdowns.

ACKNOWLEDGEMENTS

This study was supported by the Universidad CEU

Cardenal Herrera, Ford Spain S.L. and Fundaci

para el Desarrollo y la Investigaci

on (FDI), Spain,

which the authors gratefully acknowledge.

REFERENCES

Ahmetoglu, M., Broek, T., Kinzel, G., and Altan, T. (1995).

Control of blank holder force to eliminate wrinkling

and fracture in deep-drawing rectangular parts. CIRP

annals, 44(1):247–250.

Endelt, B., Tommerup, S., and Danckert, J. (2013). A

novel feedback control system–controlling the ma-

terial ﬂow in deep drawing using distributed blank-

holder force. Journal of Materials Processing Tech-

nology, 213(1):36–50.

Garcia, E. and Montes, N. (2019). Mini-term, a novel

paradigm for fault detection. IFAC-PapersOnLine,

52(13):165–170.

Garc

ıa, E., Mont

es, N., and Alacreu, M. (2018). Towards

a knowledge-driven maintenance support system for

manufacturing lines. In ICINCO (1), pages 53–64.

Hardt, D. E. (1993). Modeling and control of manufacturing

processes: getting more involved.

Hardt, D. E. (2017). Forming processes: Monitoring and

control. The Mechanical Systems Design Handbook:

Modeling, Measurement, and Control.

Hardt, D. E. and Fenn, R. C. (1993). Real-time control of

sheet stability during forming.

Hoffman, H. (1998). Metal forming handbook. Springer

Science & Business Media.

Hsu, C.-W., Ulsoy, A., and Demeri, M. (2000). An ap-

proach for modeling sheet metal forming for process

controller design. J. Manuf. Sci. Eng., 122(4):717–

724.

Hsu, C.-W., Ulsoy, A., and Demeri, M. (2002). Develop-

ment of process control in sheet metal forming. Jour-

nal of Materials Processing Technology, 127(3):361–

368.

Lim, Y., Ulsoy, A. G., and Venugopal, R. (2013). Process

control for sheet-metal stamping. Springer.

Lim, Y., Venugopal, R., and Ulsoy, A. (2012). Auto-tuning

and adaptive stamping process control. Control Engi-

neering Practice, 20(2):156–164.

Lim, Y., Venugopal, R., and Ulsoy, A. G. (2008). Advances

in the control of sheet metal forming. IFAC Proceed-

ings Volumes, 41(2):1875–1883.

Lim, Y., Venugopal, R., and Ulsoy, A. G. (2010). Multi-

input multi-output (mimo) modeling and control for

stamping. Journal of dynamic systems, measurement,

and control, 132(4).

Magraner, E. G. (2016). An

alisis de los sub-tiempos

de ciclo t

ecnico para la mejora del rendimiento de

las l

ıneas de fabricaci

on/tesis doctoral presentada

por Eduardo Garc

ıa Magraner; dirigida por Nicol

Mont

es S

anchez. PhD thesis, Universidad CEU-

Cardenal Herrera.

Maugin, G. (1980). The method of virtual power in con-

tinuum mechanics: application to coupled ﬁelds. Acta

Mechanica, 35(1):1–70.

Ng, Y.-M. H., Yu, M., Huang, Y., and Du, R. (2007). Di-

agnosis of sheet metal stamping processes based on

3-d thermal energy distribution. IEEE transactions on

automation science and engineering, 4(1):22–30.

Shang, J. and Daehn, G. (2011). Electromagnetically as-

sisted sheet metal stamping. Journal of Materials Pro-

cessing Technology, 211(5):868–874.

Siegert, K., Ziegler, M., and Wagner, S. (1997). Closed

loop control of the friction force. deep drawing pro-

cess. Journal of materials processing technology,

71(1):126–133.

Song, J., Kim, S., Liu, Z., Quang, N. N., and Bien, F.

(2016). A real time nondestructive crack detection

system for the automotive stamping process. IEEE

Transactions on Instrumentation and Measurement,

65(11):2434–2441.

Ubhayaratne, I., Xiang, Y., Pereira, M., and Rolfe, B.

(2015). An audio signal based model for condition

monitoring of sheet metal stamping process. In 2015

IEEE 10th Conference on Industrial Electronics and

Applications (ICIEA), pages 1267–1272. IEEE.

Viswanathan, V., Kinsey, B., and Cao, J. (2003). Experi-

mental implementation of neural network springback

control for sheet metal forming. J. Eng. Mater. Tech-

nol., 125(2):141–147.

Wagener, H.-W. (1997). New developments in sheet metal

forming: sheet materials, tools and machinery. Jour-

nal of materials processing technology, 72(3):342–

357.

Yagami, T., Manabe, K.-i., Yang, M., and Koyama, H.

(2004). Intelligent sheet stamping process using seg-

ment blankholder modules. Journal of Materials Pro-

cessing Technology, 155:2099–2105.

ICINCO 2021 - 18th International Conference on Informatics in Control, Automation and Robotics

402