Condition Monitoring of Paper Machine with Thermal Imaging
Infrared imaging creates new opportunities to monitor temperature and moisture occurrences in several sections of paper machine. Thermal cameras can detect occurrences and changes in these quantities that are undetectable by other measurement methods. Rapid development in the IR-technology, especially improvements in the area of uncooled bolometer cameras, enables the use of IR-cameras in continuous monitoring, even in the harsh paper mac-hine environment. Predictive condition monitoring has a central role the maintenance of paper machinery since the aim is to maximize the availability of these expensive manufacturing lines. Continuous monitoring not only prevents mechanical breakdowns and production losses but also helps operators to optimize the service lives of machine components. Until now, thermal monitoring in paper machinery has been based on portable infrared cameras and slow-speed line-scan cameras. The both are applicable to detect continuous occur-rences such as thermal or moisture stripes. However, their resolu-tion in machine direction is insufficient to detect spot-like objects. Other trend has been the use of quantum well cameras, or equiva-lent cameras with high frame rate, to detect also spot-like occur-rences. Unfortunately these cameras need efficient cooling system, which dramatically decreases the service-life of them. More significant problems involved with the high performance cameras are the prices that are typically about or more than 100,000€. Thus the cooled, high performance cameras can be used only for non-continuous troubleshooting systems with single camera. Our research has been concentrated on monitoring two rotating components of paper machine, calender polymer rolls and press fabrics. The main challenges in these applications are the high speed of the target surface, obtaining adequate temperature diffe-rence resolution and challenging environmental conditions. In this paper, synchronization methods are discussed as a solution to enhance the resolution in machine direction. The both applications and their characteristics are discussed in the following chapters, as well as some image processing techniques. Monitoring calender polymer rolls and press fabrics The polymer cover of calender rolls is vulnerable to impacts and local temperature differences, called as hot spots. Hot spots may appear if dirt or any material attaches on the roll surface and thus causes an impact on the surface every time the object goes through the nip. The repetitive impacts will heat the surface and, due to its low coefficient of heat expansion, the surface will expand. This increases the impact power and eventually breaks the polymer covering. Thermal imaging is the only reliably method to detect hot spots because the vibration levels caused by them are typically too low for being detected by accelerometers. Polymer rolls may be damaged also as a consequence of line load failure or too high surface temperatures of the thermal metal roll next to it. These failures result in thermal bands around the surface of the roll and therefore they do not lead to detectable vibrations. Local temperature differences on the roll surfaces can be detected directly by infrared cameras, whereas congested areas in fabrics can be detected indirectly due to the deviating emissivity and temperature values of the congested areas. Filling of the press fabrics caused by dirt and other contaminants will unevenly wear the fabrics, thereby shortening their service life. In addition, the water content in the congested part of fabric is higher than in those that are clean, causing problems with the paper quality. Uneven moisture profile of the paper web will also hamper the runability in the drying section. The cost of the polymer roll damage is very high because of the cost of repair or replacement of the cover and the loss of paper production. Furthermore, a breaking roll cover is a serious risk of injury to the people working close to the calender. The damage of polymer roll can develop during only a few minutes. In contrast, a fabric damage can occur over few hours and will not cause direct danger to people in the mill. Still the knowledge of the condition of the fabrics allows operators to optimize the service lives and the change sequences of the fabrics. Synchronization methods As a result of development work, two synchronization methods for imaging rotating objects have been discovered. The more efficient one, adaptive triggering, can not be used with present bolometer technology and hence the other method, based on timing images, has been developed. The adaptive triggering is an efficient synchronization method if the occurrence to be examined is continuous, repetetive and relatively high speed proportional to the frame rate of the camera. This is the case when we are imaging rotating paper machine components at full speed. With adaptive triggering, the particular areas of the object surface can be imaged by assigning the exact moment of each frame. The whole surface will be imaged during a finite number of rotations by changing the selected area for each image. A trigger sensor is used for determining the start of rotation of a roll or fabric. The sensor sends a signal when an identifier (a magnet, bolt or reflective object, depending on the sensor's type) attached on the object bypasses it. Figure 1 shows how images can be focused on certain sectors of a roll surface by adaptive triggering. The image on the left represents the areas covered by two consecutive images taken in the same rotation. These areas consist of Field Of View (FOV) and stretched image due to movement of the object surface during camera's integration time. The image on the right represents areas covered by images taken in four sequential rotations (numbered). The images are triggered so that each image places right next to the corresponding image taken in the previous rotation. The arrow illustrates the blank area that reduces in every subsequent rotation. The use of adaptive triggering requires real time triggering of the camera. Nevertheless, the present bolometers do not offer exact triggering and even most of the high performance infrared cameras can be triggered only with unknown delay. That is, the triggerable cameras will read incomplete frame at its end while they would get trigger signal during the reading. Hence the starting moments of externally triggered frames are uncertain and must be checked afterwards, as in the other synchronization method. The other synchronization method utilizes timing of the frames and the trigger signal from the sensor. In this method the camera is not triggered externally. Standard cameras send a vertical synchroni-zation signal between every frame at their normal frame rate. The timing is implemented by delay module that measures the elapsed time between the vertical synchronization signal and the trigger signal. Based on this information the frame can be focused into the object image. The disadvantages in this method are inefficiency because of the overlapping of the images and the possibility that the image will not complete in certain situations. These situations appear if the frame rate is divisible by the rotation frequency of the object roll or fabric. Then the images will place on the same areas of the surface at every rotation. Image Processing Due to the fast-developing nature of polymer roll damage and because of the generation of complete images of roll surfaces takes several rotations, the usable time for image processing is limited. The very regular temperature distribution on the polymer roll enables the use of residual calculation as a means for detecting spot-like objects that differ from the average temperature. The calculation simply compares the temperature value of each pixel with the mean temperatures in the cross and machine directions. The operation is fast enough for on-line monitoring and highlights spot-like occurrences. The residual calculation is as follows:
The mean temperature value term of complete image is used for returning the original intensity level that decreases due to residual calculation. This was the best way for us to compare residual images with original images. However, the term can be ignored or replaced with other suitable correction factor depending on the need of application. In addition, the residual calculation can be used with averaging several consequence images that increases signal-to-noise-ratio. Nevertheless, this attenuates contrast and it is advisable to use only when necessary. Conclusions Synchronization methods enable bolometer cameras to be used in continuous monitoring of rotation paper machine components. We are looking forward to see how bolometer cameras can perform in comparison to the high performance cameras in tests that will take their place in March 2002. Before infrared cameras can be imple-mented in continuous use, there are many problems to be solved, such as protection the cameras from heavy vibrations, dirt, tem-perature and humidity that are always present in paper machine environment.
Fig. 1
Contact information: Jarno Suomela
|
