Abstract

This paper presents the results of residual stress measurements on a working mill roll in cold rolling steel mill No 1 and on back-up rolls in cold rolling steel mill No 2, both in Tadeusz Sendzimir Steel Works in Cracow, Poland. Stress measurements were made with a STRESSCAN 500C instrument using the principle of Magnetic Barkhausen Noise Analysis.

Introduction

The exact choice of the steel alloy composition used for a steel mill roll, the prediction of its performance as well as the monitoring of its condition in operation are difficult and responsible tasks of metallurgists. Steel mill rolls are very expensive and thus their durability is of fundamental importance. Residual stresses in a roll are the primary factor influencing its durability. [Ref. 1]. Too high stresses or their unfavourable distribution should be a sign to change or regenerate a roll [Ref. 2]. Regeneration of a fatigued roll, by grinding off a sufficiently thick layer, may prevent a fracture of the mill and a costly interruption in the process [Ref. 3].

As the residual stresses in a steel mill roll build up continuously during a campaign it is important to know their level before the installation of a roll in its cage and after a rolling campaign.

Analysed Materials

Stress measurements were carried out on the surface of one working roll in the cold rolling mill No 1 and on the surfaces of three back-up rolls of in the cold rolling mill No 2.
The working roll, produced in Poland, had the following composition:
0.80%C, 0.40%Mn, 0,30%Si, 1.80%Cr, 0.25%Ni, 0.10%V, 0.25%Cu, 0.03%P and 0.03%S
The working roll had the dimensions of Ø=415 mm and L=1200 mm. The heat treatment process involved: austenitizing at 860°C, quenching in water and low-temperature tempering at 150°C during 100h. The backing rolls, one supplied by a Belgian manufacturer and two by Japanese manufacturers, had the dimensions of Ø=152° mm and L=1700 mm. They have been used in the cold rolling mill No 2. since three years. The manufacturers did not reveal the chemical compositions of these rolls nor the heat treatment process.

Instrumentation

Stress measurements were made using a commercially available STRESSCAN 500C instrument. It is based on the principle of Magnetic Barkhausen Noise Analysis [Ref. 4].
A more detailed description of the measurement principle is given in chapter 4.
The device allows a fast stress measurement on the surface of the rolls in operating conditions.
Measurements can be made in three nominal depths of penetration, 20µ m, 70µ m and 200µ m. The data presented here are limited to the 200µ m range which is significantly more than using the X-ray diffraction method (8 - 13µ m) and thus better representative of the surface layer. Typical repeatability of the Barkhausen noise intensity - MP (Magnetoelastic Parameter) in measurements on the same location on the sample surface by removing / replacing the sensor has an error of about 2%.

Magnetic Barkhausen Noise Analysis

The Magnetoelastic Barkhausen Noise technique is based on a relatively simple concept that was discovered back in 1919. Ferromagnetic materials consist of small magnetic regions resembling individual bar magnets called domains. Each domain is magnetised along a certain crystallographic easy direction of magnetisation. Domains are separated from one another by boundaries known as domain walls. Applied magnetic field will cause domain walls to move. In order for a domain wall to move, the domain on one side of the wall has to increase in size while the domain on the opposite side of the wall shrinks. The result is a change in the overall magnetisation of the sample.

If a coil of conducting wire is placed near the sample while the domain wall moves, the resulting change in magnetisation will induce an electrical pulse in the coil. The first electrical observations of domain wall motion was made by professor Barkhausen in 1919. He proved that the magnetisation process, which is characterised by the hysteresis curve, in fact is not continuous, but is made up of small, abrupt steps caused when the magnetic domains move under an applied magnetic field. When the electrical pulses produced by all domain movements are added together, a noise-like signal called Barkhausen noise is generated.

Barkhausen noise has a power spectrum starting from the magnetising frequency and extending beyond 2 MHz in most ferromagnetic materials. It is exponentially damped as a function of distance it has travelled inside the material. This is primarily due to the eddy current damping experienced by the propagating electromagnetic fields that domain wall movements create. The extent of damping determines the depth from which information can be obtained (measurement depth). The main factors affecting this depth are

1. frequency range of the Barkhausen noise signal analysed, and
2. conductivity and permeability of the test material.
Measurement depths for practical applications vary between 0.01 and 1.5 mm.

Two important material characteristics will affect the intensity of the Barkhausen noise signal. One is the presence and distribution of elastic stresses which will influence the way domains choose and lock into their easy direction of magnetisation. This phenomenon of elastic properties interacting with domain structure and magnetic properties of material is called a "Magnetoelastic interaction". As a result of Magnetoelastic interaction, in materials with positive magnetic anisotropy (iron, most steels and cobalt), compressive stresses will decrease the intensity of Barkhausen noise while tensile stresses increase it. This fact can be exploited so, that by measuring the intensity of Barkhausen noise the amount of residual stress can be determined. The measurement also defines the direction of principal stresses. The other important material characteristic affecting Barkhausen noise is the metallurgical structure. This effect can be broadly described in terms of hardness: the noise intensity continuously decreases in microstructures characterised by increasing hardness. In this way, Barkhausen noise measurements provide information on the microstructural condition of the material.

Extensive calibration tests have shown [Ref. 5] that the Barkhausen noise response extends over the full elastic range in most steels, with some levelling off characteristic of the "S" shape at the tails of the curve as shown in figure 1. Hard martensitic steels show an early saturation at the compressive end of the calibration curve.

Fig 1: A typical calibration curve. Barkhausen Noise Intensity (MP) vs. Stress

In its non calibrated mode, Stresscan 500C output is a Magnetoelastic Parameter called MP, proportional to the peak level of measured Barkhausen noise. Through a calibration, using strain gages and test samples made of the same material as that to be analysed, measured MP values are converted to absolute stresses, expressed in MPa. Calibration curves are obtained by measuring the level of Barkhausen noise while stressing the test piece in compression and tension. For strain measurements a commercially available strain indicator and miniature strain gages are used. Strain data are converted into stresses using formulae derived from the Hook's law of elasticity. MP vs. stress calibration data are stored in the memory of the microprocessor controlled Stresscan 500C device.

The Object and the Scope of the Test Method

Residual stresses in the surface of the working roll (No 450) were measured at the following stages of its exploitation:

1. upon the delivery from the manufacturer
2. after the first campaign, in which 350 tons of steel were rolled
3. after grinding off the deteriorated layer of 0.3 mm
4. after the second campaign during which a cobble (sticking of the rolled sheet to the roll) occurred.

NEW ROLL - BEFORE THE FIRST CAMPAIGN

AFTER THE FIRST CAMPAIGN

AFTER GRINDING OFF 0,3 mm

AFTER THE SECOND CAMPAIGN

Fig 2: Axial and hoop stress distribution in working roll No 450 (diameter 415) after each stage of the exploitation in a cold rolling mill No 1

Residual stresses in the backing roll (No 37) from the first manufacturer were measured at the following stages:

1. after grinding (referred to as first grinding), where 1.0 mm layer was removed
2. after a campaign (referred to as first campaign), in which 40607 tons of steel were rolled
3. after subsequent grinding (referred to as second grinding), where 0.55 mm layer was removed

Residual stresses in the backing rolls (No 11 and 19) from the second manufacturer, working simultaneously in the same mill cage, were measured after the campaign ( referred to as first campaign ) during which each of them processed 33688 tons of steel. Roll 11 was used as the upper-, while roll 19 was used as the lower roll.

The measurements carried out in this project were intended to obtain information about the change in the level of residual stress in the rolls supplied by various manufacturers, in function of the quantity of rolled steel as well as the influence of the location of the roll in the cage.

All measurement points were precisely defined along a generating line, in distances of 100 mm from each other. Precise definition of the measurement points allowed the subsequent measurements to be made in exactly the same locations after each campaign and grinding. Measurements in the direction parallel to the axis of the roll (axial) and perpendicular to the axis (hoop) were made in each point.

Measurement Results and Discussion

The results of residual stress measurements in the working roll after each stage of its exploitation are shown in the figure 2.

It can be seen that in the new roll the stresses are highly compressive. Along almost the full length of the roll the stress values are uniform. Only towards the edges the stress values rise (become less compressive).

After the first campaign the hoop stresses rose further, becoming less compressive. This could be explained as fatigue of the surface layer due to exposure to cyclic loading in tension and compression.

After grinding off a layer of 0,3 mm i.e. the thickness spoiled during the preceding campaign the measurements showed that the surface stress level did not return to the state prior to the first campaign. This indicates that grinding off a layer of 0,3 mm was not sufficient.

A significant difference between the axial and hoop stresses, similar to that after the first campaign, supports the conclusion that the grinded off layer of 0,3 mm was inadequate. Furthermore, a local increase of surface stresses (less compressive values) both axial and hoop, was recorded in the same location where equally high values were measured directly after the campaign.

AFTER GRINDING OFF (I) A LAYER OF 1,0 mm

AFTER THE CAMPAIGN (I)

AFTER GRINDING OFF (II) A LAYER OF 0,55 mm

Fig 3: Axial and hoop stress distribution in backing roll No 37 (diameter 1520 mm) after each stage of the exploitation in a cold rolling mill No 1

During the second campaign a cobble occurred. In the areas of material sticking both stresses, axial and hoop) rose radically. This effect could have been strengthened by any change in the microstructure of the surface layer. It is well worth noticing that one of the two cobbles took place there where locally higher residual stresses were measured before and after the grinding of the 0,3 mm layer. Although it is primarily the poor quality of the rolled material that will result in a higher risk of cobbling, it can be postulated that a high local surface stresses will encourage this type of failure.

The results of residual stress measurements in the backing roll after each stage of its exploitation are shown in the figure 3.

It can be seen that after grinding off (I) the layer of 1.0 mm the surface stresses close to the edges of the roll are slightly higher (less compressive) than in the middle. Nevertheless they are uniform, both in the axial and hoop directions. This seems to indicate that grinding off a thickness of 1 mm was sufficient to remove the spoiled layer after the preceding campaign.

After the campaign (I) (see Fig. 3.) during which 40607 tons of steel were rolled, surface stresses in the axial direction almost did not change, while the hoop stresses rose radically.

After grinding off a layer of 0,55 mm the surface stresses measured were practically the same as before the campaign. It can be concluded that the removal of the layer of 0.55 mm was adequate.

Remarkable is also a comparison of the magnitude of the measured residual stresses induced during the campaign in rolls from different suppliers as well as after various quantities of processed steel and working in different locations.
Roll No 37, from the first supplier rolled 40607 tons of steel whereas rolls 11 and 19, from the second supplier worked both in the same cage and rolled 33688 tons of steel. Roll 11 worked as the upper- and roll 19 as the lower roll. The results of residual stress measurements after such exploitation are presented in the figure 4.

It can be seen that in all three rolls axial stresses were all over slightly compressive. Only in the zones close to the edges, also after the grinding, the measured stresses were slightly higher (tensile). The hoop stresses in all three rolls after the campaign were considerably higher. A particularly high rise in the measured stress level after the campaign was found in roll 37.

The results of the measurements shown in the figure 4. indicate that the quantity of rolled steel has a prominent influence on the state of the surface stresses. A distinct influence can be ascribed to the location of the roll in the cage. Based on the presented results it may be concluded that that the lower roll is exposed to tougher conditions than the upper one. It is also apparent that the manufacturing process of a roll has an influence on how sensitive or insensitive it is to surface stress build-up during a rolling campaign.

BACKING ROLL No 11

BACKING ROLL No 19

BACKING ROLL No 37

Fig 4: Axial and hoop stress distribution in backing rolls No 11, 19 and 37 (diameter 1520 mm) from different suppliers after their exploitation in cold rolling mill No 2

Conclusions

It was shown that a strong increase in the surface residual stresses in a roll is observed during a rolling campaign. Initially compressive stresses reduce to 0 and subsequently tensile stresses build up, predominantly in the hoop direction.
Grinding off the spoiled surface layer restores the stress condition to that from before the campaign. It is essential to entirely remove the layer damaged during previous campaign.

Systematic control of the stress levels in a roll, using a Stresscan 500C each time after rolling a determined quantity of steel helps optimise the exploitation time of a roll. It may also help to differentiate between various types of rolls from different suppliers and predict their sensitivity to stress build-up during a campaign.

Measuring the stresses using the Stresscan 500C after each surface layer removal by grinding helps to establish if the spoiled roll surface was adequately removed. This in turn helps increase the number of rolling campaigns of the roll and results in an increase in the productivity of the mill.

References

1. Santoli Pat A., "Hot Strip Mill Backup Roll Failure: Stress and Metallurgical Analysis" Allegheny Ludlum Steel Corporation, pp.225-237
2. Corbett Robert B., "Rolls for the Metalworking Industries" AIME Iron and Steel Society
3. Tiitto Kirsti, "Measuring Stresses in Rolls by Magnetoelastic Method" The 27th Mechani- cal Working and Steel Processing Conference, Cleveland, Ohio, October 27-29, 1985, pp.273-279
4. Tiitto Kirsti, "Use of Barkhausen Effect in Testing for Residual Stresses and Material Defects, Non-Destructive Testing - Australia, Vol.26 No.2 March/April 1989 pp. 36-41
5. Tiitto Seppo and Wojtas Andrzej, "Magnetoelastic Testing of Biaxial Stresses" VDI Berichte Nr. 882, 1991, pp. 291-301