Electromagnetic sensors for in-situ dynamic microstructure monitoring of recovery and recrystallisation in interstitial free steels

An in-situ high temperature EM sensor was used to detect changes in the microstructural state caused by recovery and recrystallisation for IF steel samples during heat treatment. The relative permeability values determined from the sensor measurements at the different heat treatment temperatures showed the expected strong influence of temperature on the magnetic measurement. The relative permeability values determined at each temperature increased during the recovery and recrystallisation processes and correlated well with the different microstructural states determined by EBSD. The results show that the EM sensor measurement approach shows significant potential for in-situ microstructure monitoring during the annealing process of IF steels, with larger magnetic parameter changes between microstructural states at the higher temperatures typical of commercial annealing practice. In addition it is seen that the magnetic, and EM sensor, measurements show more sensitivity to recovery than hardness measurements, and can also monitor recrystallisation.


Introduction:
Interstitial-free (IF) steels are widely used in the automotive industry due to their excellent formability, which secures a high stamping success rate leading to cost reduction and weight saving. The high formability of IF steels is due to their low yield strength, high ductility and high r-value, achieved by the formation of desired crystallographic texture (high fraction of fibre {111} texture component). The excellent mechanical properties are achieved through the processing schedule and by minimising the interstitial elements (C, N) present in the chemical composition and using stabilising elements such as Ti and Nb, which can further eliminate the interstitial elements by forming carbides [1,2]. The annealing process post-cold rolling, at typical temperature range of 590 -700°C, is particularly important in achieving the desired final texture and grain size through recrystallisation. Online mathematical based modelling systems are applied to optimise the processing parameters and predict the microstructure from the temperature-time profile in the process line for known cold rolling reductions and compositions [3][4][5]. Knowing when recrystallisation occurs and is completed would significantly benefit optimisation of the continuous annealing lines for efficiency. Electromagnetic (EM) sensors have shown great potential in characterising steel microstructure both offline and online. EM systems (IMPOC, HACOM and 3MA) are routinely deployed online, providing information on microstructure and strength levels for cold strip steels [6][7][8]. For online microstructure control, the EMspec TM system has emerged for transformation monitoring during hot strip steel processing [9]. The operating temperature and harsh environmental conditions for high-temperature in-situ measurement mean that sensors like the EMspec TM system operate at a relatively large lift-off, resulting in a low applied magnetic field at the target sample. Whilst the EMspec TM sensor monitors hot strip steel, it operates at room temperature due to the significant lift-off and its water-cooled canister. The EMspec TM sensor signal depends on the low magnetic field relative permeability and electrical conductivity of the steel sample, which is related to microstructural features such as austenite to ferrite phase transformation. Recent reports have shown the feasibility of using induction spectroscopy to detect transformation in hot rod or strip samples (where the sensor is at room temperature) in the laboratory [10,11], although the monitoring of any specific microstructure changes was not fully reported and these sensors are at room temperature.
Magnetic parameters are affected by the different microstructural features in steel (e.g. phase fraction, grain size, precipitates, dislocation density). EM sensors in the laboratory have been shown to have sensitivity to these factors [12][13][14][15]. The magnetic properties also change significantly with temperature and there is limited data available for high temperature magnetic behaviours. The low field relative permeability and BH curve changes with temperature for pure iron, and a limited number of structural steels have been studied up to the Curie point using ring samples or high-temperature cylindrical EM sensors [16,17].
The annealing process after cold rolling for IF steels comprises three different metallurgical stages of recovery, recrystallisation and grain growth. During annealing, the changes in microstructural parameters include dislocation density (from the cold rolling plastic strain) and distribution, crystallographic texture and ferrite grain size. Thompson et al. reported that the initial relative permeability decreases dramatically with increased plastic strain up to 2 -3% for ferrite -pearlite steels due to the increased pinning sites of domain wall motion at the dislocation tangles resulting from the deformation process [18]. The pinning effect of dislocations becomes less effective when the dislocation density in the steel continues to increase, reaching a saturation value; hence further increase in strain beyond 3% has little effect on the relative permeability values [18]. After the recrystallisation process, specific grain size and texture are required to achieve the desired mechanical properties through the grain growth stage. It is known that the magnetic properties of steel are also affected by the ferrite grain size. Thompson et al. reported an increase in initial and maximum permeability with an increase in grain size from 25 -60 µm in a 0.17 wt% C steel sample [18]. Zhou et al. reported the low magnetic field (~7 A/m) relative permeability changes from 80 to 327 for a grain size change of 2.5 μm to 52 μm in a single-phase ferritic microstructure. Also, the permeability changes from 171 to 211 for a grain size change from 13 μm to 63 μm in a dualphase (70% ferrite and 30% pearlite) microstructure. The strong correlation between grain size and magnetic properties is believed to be due to the magnetic domain structure and domain wall motion being affected by grain boundaries [19]. The grain boundaries are considered obstacles to magnetic domain wall motion (due to the higher dislocation density around it and the likelihood of forming closure domains) [12,[20][21][22]. Some research has been carried out to study the effect of recovery and recrystallisation on the full hysteresis magnetic properties in plastically deformed low carbon, IF and electrical steels [23][24][25][26][27]. It was reported that parameters derived from magnetic hysteresis loops such as coercivity, maximum permeability, remanence and losses are closely correlated to recovery and the onset of recrystallisation. The change in the hysteresis loop magnetic parameters is believed to be due to the annihilation of dislocations during the recovery process [28] while the grain size remains the same. Theoretical predictions indicate that coercivity is proportional to the square root of the dislocation density [29]. However, during recrystallisation, the grain size reduction can increase coercivity and, therefore, compensate for the change in coercivity caused by dislocation annihilation during recrystallisation [15][16][17]24]. Gurruchaga et al. reported that remanence Br could be used to characterise the recrystallised fraction in low carbon steel when a significant weakening in the intensity of the α fibre texture components and the enhancement of the γ fibre texture components happens.
To date, there are no reports on how magnetic properties change in-situ during recrystallisation at high temperatures. This paper includes room temperature low field EM sensor measurements for interrupted heat-treated IF steels samples, showing the recovery and recrystallisation stages agreeing with literature reports, and in-situ high temperature measurements for dynamic changes of signal with recovery and recrystallisation progression. The high-temperature cylindrical EM sensor used is capable of operating at high temperatures inside a furnace. The in-situ EM sensor measurements during annealing heat treatments at a range of temperatures from 365 -700°C for IF steel grade are reported and the calculated relative permeability values have been correlated with microstructural parameter changes.

Materials and methods
A commercial IF grade steel micro alloyed with Ti was supplied in as 1 mm thickness coldrolled strip. Samples for microscopy were polished to an OPS finish and etched in 2% nital. The EBSD micrographs were taken using a JEOL JSM-7800F scanning electron microscope with AZtech Oxford Instrument software. The hardness values were measured with an Indentec 5030 SKV Vickers hardness testing machine with a 5kN load.
For continuous in-situ EM sensor measurements, IF steel samples measuring 110 mm x 19 mm x 1 mm, were prepared by cutting rectangular pieces from the parent sheet, the longest side of the sample aligned with the rolling direction. A high-temperature cylindrical EM sensor, capable of measuring up to 900 o C, was used for the EM measurements. The exciting and sensing coils are air-cored and formed around a ceramic cylinder, with the coils, made from K-Type thermocouple wires, wound concentrically on the cylinder. The sensing coil has 54 turns, and the exciting coil has 50 turns. A schematic diagram of the high-temperature cylindrical sensor, with dimension details, is shown in Figure 1. A Solartron 1260A impedance analyser drove the coils, and the real inductance values were determined from mutual inductance measurements. The sensor was placed inside a laboratory muffle furnace and allowed to stabilise at the furnace set temperature before the sample was placed inside the sensor. When the strip sample is inserted into the sensor, the EM measurement starts. K-type thermocouples spot welded on the steel were used to record sample temperature. The time for a sample to reach a target annealing temperature for each measurement was determined and was typically under two minutes. The sensor output readings were taken at a single frequency of 100Hz every two seconds for off-line and in-situ EM sensor measurements. This relatively low value of frequency was chosen to give minimal eddy current effect so that the sensor output signal is dominated by the relative permeability rather than the conductivity of the sample; whilst also giving a stable signal and high signal to noise ratio, as the induced voltage in the sensing coil increases with frequency. The measured real inductance values are directly related to the magnetic low field permeability of the sample. A finite element model was developed for the high-temperature cylindrical sensor. The geometry and details of the sensor/sample model were the same as the experimental setup. The sensor model calibration was performed by fitting the sensor model to sensor measurements for reference samples of known low field relative permeability and resistivity. Close fit (less than 1% error) between the modelled and measured real inductance values for all the samples have been achieved. Details of the EM sensor output models were described in [30] and [31]. Then, the low field relative magnetic permeability values were determined by fitting the modelled inductance with the experimentally measured ones based on a non-linear least square method in COMSOL Live Link for MATLAB. Details of the fitting method can be found in [13].

Results and discussion
In-situ EM sensor measurement for the cold-rolled IF sample during annealing, involving heating and holding at 700°C, was carried out. After the annealing treatment, where full recrystallisation and grain growth was achieved, the same sample was tested again in the insitu EM test to determine the effect of temperature on the EM sensor readings in the absence of microstructural changes. The measurement results are shown in Figure 2. It can be seen that the two samples show different initial inductance values due to the different microstructure states, one being cold-rolled and the other was fully annealed. For both samples, the real inductance increases with time initially due to the known effect of temperature on the low field permeability of steels [30,32]. It is also possible that some recovery may occur during the cold-rolled sample's heating stage to 700°C. Thermocouple measurements showed that the time for the steel sample to reach the target heat treatment temperature was approximately 120 secs (indicated by the green dashed line in Figure 2). Therefore, the changes in inductance after that point are expected to be due to microstructural changes only. The cold-rolled sample undergoes recovery and recrystallisation, which causes a significant increase in inductance values. In comparison, the re-heating of the annealed sample shows little change in inductance value from 120 secs to 10k secs. Eventually, the two samples reached a similar inductance value after around 1600 sec. Figure 3 shows the in-situ EM sensor measurement results for the cold-rolled and a fully annealed IF sample heated to 420°C. The time to reach the furnace temperature was about 120 secs, marked by the green vertical dashed line in Figure 3. It is expected from literature that at this temperature only recovery should have taken place [27]. Similar to the results for heating to 700°C, the initial increase before 120 sec is due to the known temperature effect on permeability. The continuing increasing signal for the cold-rolled sample after 120 sec is believed to be due to recovery, which has been reported to increase the relative magnetic permeability values when room tempearature magnetic measurements are made for cold rolled and for recovered IF samples [15][16][17]24]. However, the annealed sample shows only negligible change in inductance with time after 120 sec when the sample reaches 420°C, indicating that little microstructural change occurs.  To determine how sensitive the EM sensor is to the microstructural state and rate of recovery and recrystallisation, in-situ EM sensor measurements were carried out on cold-rolled IF steel samples annealed at 365 o C, 420 o C, 650 o C and 700 o C. It can be seen from Figure 4

Annealed sample
Cold rolled sample EM sensor response for each of the annealing temperatures shows different rates of change in inductance with time. The inductance values at the end of the heat treatment hold time are significantly different because the samples are at different temperatures, and temperature has a large effect on the magnetic signals, as well as the steel being in different microstructural states. At 365 o C and 420 o C, only recovery is expected to have taken place, as literature suggest that only recovery is observed at the annealing temperatures of <600 o C [27]. The inductance curve for the 420 o C annealing heat treatment shows a slightly faster rate than the 365 o C annealing heat treatment to reach the signal plateau, which could be due to the higher heating rate (when you put a sample into the hotter furnace) and the fact that permeability will be higher at higher temperature. Hardness results were 191±2 and 192±2 Hv for the 365 o C and 420 o C samples after annealing (samples quenched after removal from the EM sensor in the furnace), respectively. The change in hardness is very small compared to the as-received cold-rolled sample hardness of 200±2 Hv. The hardness values are consistent with the reports by Martinez-de-Guerenu et al., who stated that hardness is relatively insensitive to microstructural changes due to recovery [33]. In contrast, the hardness values dropped to 84±2 and 82±2 Hv for the fully recrystallised samples, obtained at the end of the 650 o C and 700 o C annealing heat treatments respectively. It can be seend from Figure 4 that the shape of the inductance vs time curves for the 650 and 700°C samples has show two different stages (after reaching temperature) which might be linked to the different microstructural changing stage if recovery followed by recrystallisation. The higher gradient occurring at the shorter time for the 700°C sample may have indicated that recrystallisation occurring earlier than the 650°C sample. In order to use the results and understanding obtained from these sensor measurements in the design of new sensors, such as robust systems that can be used in industrial plant, the change in magnetic properties, not just one type of sensor measurement value, are needed. To translate the EM sensor measurement of inductance (which is affected by the sensor design and sample size) to the magnetic property (low magnetic field permeability) of the material an FE model for the sensor-sample was used, as described in the experimental section. The relative permeability change with heat treatment time for the cold rolled samples annealed at 365 o C, 420 o C, 650 o C and 700 o C are shown in Figure 5.
The relative permeability values at the end of the annealing heat treatment are different mainly because of the temperature effect. The determined permeability values for the IF steel at the annealing temperature in comparison with literature reported pure iron sample [16] are shown in Figure 6. It is interesting to note that the releative permeability value for the 700°C annealed IF steel is the same as for the pure iron sample, which might be expected due to the similarity in structure (both are 100% ferrite). The lower permeability of the 650°C annealed IF sample is likely to be relate to the fine grain size of 7.4±5 µm resulting from recrystallisation at a lower temperature. The grain size of the 700°C annealed IF steel (19.2±7.2µm) is also significantly smaller than the pure iron sample (155±68.1 µm). However, it is known that the permeability values are affected significantly by grain size in the range of 2.5 -20 µm, in compare with larger ones (>20 µm) [34]. The lower permeability values for the IF steel at lower temperatures (365°C and 420°C) will be due to the fact that these samples are only recovered not recrystallised and therefore still contain a high dislocation density.
It can be seen that there are several gradient changes in the increase in permeability during the annealing process for the 650°C and 700°C annealed samples in Figure 5. To analyse if these changes in gradient can be used to separate the recovery and recrystallisation processes, interrupted heat-treated samples taken for EBSD analysis at key gradient changing points (3, 40, 60 and 240 minutes at 650°C and 3, 30 and 210 minutes at 700°C) were taken, as shown on Figure 5. The EBSD micrographs for the interrupted test samples annealed at 700°C and 650°C are shown in Figure 7 and Figure 8, respectively.  For the 700°C annealing heat treatment, there appears to be a few very small equiaxed grains present in the micrograph in Figure 7A, which could be some initiation of recrystallisation. The hardness value was 181±2 (samples quenched after removal from the furnace) at this point. The change in hardness is small compared to the as-received cold-rolled sample hardness of 200±2 Hv, which would be consistent with some recovery and very limited recrystallisation. After 30 minutes, where the change in permeability gradient is seen, more small recrystallised nuclei / grains can be observed in Figure 7B. Therefore there has been some more recrystallisation between points A and B. This is also evidenced by the hardness result which was 90±2 at this point. It was reported that a sharp drop in hardness corresponds to the onset of recrystallisation [27]. After 210 minutes, Figure 7C shows equiaxed grains, indicating full recrystallisation has occurred, possibly with some grain growth. The hardness values further dropped to 84±2.
On the other hand, annealing at 650°C shows no significant change in microstructures after 3 and 40 minutes (shown in Figure 8D and Figure 8E). The increase in permeability values is believed to be mainly due to the effect of recovery with possibly a very small fraction of microstructure that has started to recrystallise. Hardness results were 181±2 and 168±2 Hv after annealing at 650°C for 3 and 40 minutes, respectively. Ye et al. reported that the recrystallisation starting time for an IF steel is around 3-5 mins and the full recrystallisation time was 30-50 mins at this temperature [2]. After 60 minutes, after the permeability gradient change, recrystallisation has clearly started as equiaxed grains can be seen in Figure 8F. After 240 minutes, Figure 8G shows fully equiaxed grains, which indicates completion of recrystallisation; the grain size is smaller than seen in Figure 7C for the fully recrystallised sample annealed at 700°C suggesting some grain growth has occurred at the higher temperature.
It is worth noting that several aspects can affect the accuracy of the EM sensor measurement and the determination of the permeability values in this study. Firstly, during the heating process, the resistivity of the wires in the coils increases with temperature. As the impedance analyser drove the coils with fixed the current level. The applied field strength and therefore the measured inductance will not be affected. Secondly, the steel sample will undergo thermal expansion. The maximum dimension change which is at the highest temperature is 0.82mm, 0.14mm and 0.007mm in length, width and thickness respectively. This can result in a slight increase in the inductance value (up to approx. 2%) and an overestimation of the permeability values for the samples (up to approx. 5%). A second test was carried out for each sample and it was found that the same trend can be found and the tests are repeatable. The inductance value has up to 3% variation, which can also be due to the small difference in microstructure in the samples.
The results show that the low field relative permeability and, therefore, the in-situ real inductance measurement at 100Hz is sensitive to both recovery and recrystallisation. The EM sensor used in this study applies a very low magnetic field of < 50A/m, determined from the FE model [13]. At such a low magnetic field, the magnetisation process is dominated by reversible domain wall motion. The high dislocation density in the cold-rolled sample rearranges from a more uniform distribution in the grains to sub-grain boundary structures during the recovery process. This process will remove some 180° domain wall pinning sites (i.e. dislocations) and increases the mean free path for domain wall movement, increasing the permeability. The dislocation density is then significantly reduced during the recrystallisation process, and new strain-free grains are formed, followed by some grain growth. The decrease in dislocation density and grain growth have a positive impact on the relative permeability values, which agrees with offline (room temperature) coercivity and maximum permeability measurement reported in the literature [15][16][17]24].
A significant finding from the work is that the change in relative permeability values due to the recovery and recrystallisation processes are very large at high temperatures (650 -700°C), for example approximately 600 to 4500 at 700°C. This is in a similar range to that seen for transformation from austenite to ferrite at 700°C (approximately 70% to 100% ferrite fraction) [30] which can be monitored using a commercial sensor (EMspec TM sensor system) on a hot strip run out table, albeit with the sensor operating at room temperature [35], which indicates the potential for a deployable system that could be used in commercial annealing lines.

Conclusions
In-situ EM sensor measurements have been used to detect changes in the microstructural state caused by recovery and recrystallisation for IF steel samples during high temperature annealing heat treatments. The low field relative permeability values have been determined from the in-situ sensor real inductance signals, using an FE sensor-sample model. The low field relative permeability, and sensor inductance, values correlated well with the different microstructural states (cold rolled, recovered, partially recrystallised and recrystallisedevaluated by EBSD), allowing the microstructure evolution, due to the recovery and recrystallisation processes, to be monitored. The low field relative permeability values are more sensitive to recovery than hardness measurements. The study shows the potential of insitu microstructure monitoring during the annealing process of IF steels using EM sensor approaches.