Third roller

A third roller was also trialled, and again this roller was predominantly at, with a variation of under 15 μm along its length measured using the contact sensor. The measured prole of this

(a) 200 400 600 800 1000 0 5 10 15 Position (mm) Change in Diameter ( µ

m) Eddy Current Probe

Contact Sensor (b) 200 400 600 800 1000 −10 −5 0 5 Position (mm) Difference in Profiles ( µ m)

Figure 9.17: (a) The prole of the third roller, with the prole measured by the eddy current probe in blue, and the prole measured by the contact sensor in red, and (b) the dierence between the two prole measurements.

contact sensor data is shown in red. The dierence between the two measurements is shown in gure 9.17 (b), and demonstrates good agreement between the two measurement systems. The majority of the eddy current prole measurement lies within 5μm of that from the contact sensor,

with the only substantial deviation occurring at the end of the roll, where there was a signicant band of surface magnetisation. For the overall variation of approximately 15μm in this roller, that

deviation appears to be quite a large variation, but when considered in the context of larger prole variations over the range of 2 mm which is possible on the rollers these variations are negligible, particularly when a smoothing polynomial is applied to the data.

9.4 Conclusions

In order to asses the performance of the measurement system it must be compared to the industrial specication given in section 7.1. The resolution of the technique driving the coil at resonance has been demonstrated at less than the required resolution (0.9μm with greater than 90 % reliability)

over a much larger range (10 mm) during the laboratory based trials. The performance on curved samples has also been demonstrated, as has the insensitivity to transverse misalignment. The presence of water or oil contamination on the surface of the sample has been shown to have no eect (section 9.3.2), while only very large scratches have been shown to aect the eddy current measurement (section 9.3.3). Operating the coil at resonance minimises the calibration required, while section 9.2.2 shows the minimum number of calibration points necessary in order to accurately measure the lift-o over the operating range. Although the operating temperature of the measure- ment hasn't been demonstrated over the full range, the electronics should operate consistently in this range, meaning the measurement will be unaected, provided that the temperature variation throughout the course of the scan is minimised. The weight of the probe, and other electronics

which are mounted on the calliper arm shown in gure 9.10 weigh only 600 g, considerably less than the weight required by the specication.

Based on the successful demonstration of the performance in the laboratory the measurement system was trialled in an industrial environment. The performance of the measurement system was very close to that of the existing contact sensors, particularly on the roundness measurements. However, the surface magnetisation of the roller and the changes in materials properties along the length of the roller had a relatively large impact on the eddy current measurement. While the surface magnetisation is relatively easy to detect, and can be corrected to some extent by degaussing, the change in the material properties is indistinguishable from changes in the lift-o, and resulted in fairly large discrepancies, roughly 20 μm, between the two measurement systems.

Although these eects on the lift-o due to the material properties appear to be present in the industrial samples it was not possible to quantify the degree of change in the material properties in the material along the length of the roller, or to relate that to the erroneous change in lift-o that it would produce, and this would form an important part of any further work.

It should be noted that the industrial contact sensor to which the eddy current probe was compared in section 9.3 has a number of limitations. The apparent noise in the contact sensor data in section 9.3.4, which does not appear to be present in the other measurements, is only absent from the measurements in sections 9.3.1 and 9.3.3 due to the comparatively low sampling rate of the contact sensor in those measurements. While this downsampling of the contact sensor data may obscure the noise inherent in the measurement, it does not remove its eects.

Chapter 10

Prole measurements using eddy

currents: Conclusions

This chapter summarises the work presented in chapters 7 to 9 on the use of eddy current probes as lift-o sensors. The design of the prototype device, the achieved measurement performance and how well it ts the required criteria are discussed. Avenues of additional research are also proposed.

10.1 Laboratory performance

The performance of the measurements in the laboratory on at samples was encouraging, with the single frequency achieving a lift-o measurement reliability of greater than 90 % for 0.9 μm

resolution when using the phase data, across all three industrial samples and at lift-os of up to 2 mm (section 8.2). The sensitivity of the measurement to transverse misalignment, however, was higher than the desired standard, as the required ± 1 mm transverse misalignment range caused a perceived change in the lift-o of 8.1μm (section 8.3). The alternative measurement technique,

which drove the eddy current coils at resonance, had a greatly improved performance, and the reliability of the resonant frequency measurement was in excess of 97 % at lift-os of up to 3 mm across all three industrial samples. The lift-o performance was tested further on a single industrial sample, and the frequency measurements with a resolution of 0.9μm were shown to be at least 90 %

reliable even at lift-os of up to 10 mm (section 9.1). The sensitivity of the resonance measurement to the transverse misalignment of the coil was also improved, with the transverse range of ± 1 mm causing a perceived change in the lift-o of 1.0μm (section 9.2.1).