1 Introduction 1.1 Application: high pressure grinding rolls Continuous grinding and milling are processes of major importance in many industries, e.g. mining, steel or cement industry. Since their development in the middle of the 1980’s, high pressure grinding rolls (HPGR) have become key aggregates for the comminution of medium hard to brittle materials such as minerals, slag or clinker in cement industry [1], [2]. Two cylindrical rolls, one fixed and one moveable, rotate towards one-another, with diameters between 0.5 and 2 m in diameter, and lengths from 0.3 to 1 m. In particular, HPGR can process up to 2,000 t of material per hour. The pressing forces, ranging from 400 to 30,000 kN are applied by hydraulic cylinders and transferred to the material bed by the moveable roller. The high pressure in the grinding bed, from 100 to 500 MPa, exposes the particles to heavy compressive stress. This makes the process very effective, especially by reducing energy consumption compared to other milling technologies [1], [3], [4]. Aus Wissenschaft und Forschung 35 Tribologie + Schmierungstechnik · 66. Jahrgang · 4/ 5/ 2019 DOI 10.30419/ TuS-2019-0022 Verschleiß von Kernkomponenten in der Schwerindustrie ist von großem ökonomischem Interesse. Um vorzeitiges Tauschen von Verschleißkomponenten zu vermeiden, ist deren aktueller Zustand von großer Wichtigkeit für den Instandhaltungsingenieur, insbesondere wenn kein Anlagenausfall riskiert werden darf. Deshalb wird in dieser Arbeit die Anwendbarkeit von Online-Verschleißsensoren im rauen Umfeld einer Gutbett-Walzenmühle (HPGR) untersucht. Ultraschall-Distanz- und Laser-Triangulationssensoren wurden angewendet um Signale im Feld bei verschiedenen Betriebszuständen der HPGR aufzuzeichnen. Die Datenanalyse zeigte, dass beide Sensorvarianten geringe Fehlerraten (<10 %) während des Walzenbetriebs aufweisen. Der Laser-Triangulationssensor war in der Lage die Oberfläche detailreich aufzulösen. Staubkontamination war in beiden Systemen vorhanden und muss bei einem Langzeiteinsatz der Sensoren beachtet werden. Schlüsselwörter Abrasion, Verschleißsensor, Instandhaltung, Online- Messung, Condition Monitoring, Staubkontamination Wear of key components in heavy industry plays an important economic role. In order to avoid premature exchange of wear components, their current condition is of major importance for maintenance engineers, especially when no catastrophical failure can be risked. Therefore in this work the applicability of online wear sensors in the severe environment of a high pressure grinding roll (HPGR) will be studied. Ultrasonic and laser triangulation distance sensors were applied and signals recorded for different operating conditions of the HPGR. After data analysis, it was found that both sensor principles entailed low failure rates (<10 %) during operation. The laser triangulation sensor was capable of resolving the surface in great detail. Dust contamination was found at both systems and needs to be taken into account for long-term application of the sensors. Keywords abrasion, wear sensor, maintenance, online measurement, condition monitoring, dust contamination Kurzfassung Abstract Online wear measurement in harsh environment. Part 2: Application roller press Markus Varga, Reinhard Grundtner, Alexander Maurer, Martin Kirchgaßner* Eingereicht: 10. Mai 2019 Nach Begutachtung angenommen: 17. Juni 2019 * Dr.mont. Markus Varga, MSc orcid-iD: https: / / orcid.org/ 0000-0001-8272-4122 Ing. Reinhard Grundtner, MSc orcid-iD: https: / / orcid.org/ 0000-0002-2177-4879 Alexander Maurer, BSc AC2T research GmbH, Wiener Neustadt, Austria DI. Dr. tech. Martin Kirchgaßner Castolin GmbH, Wiener Neudorf, Austria TuS_4_5_2019.qxp_T+S_2018 23.08.19 13: 15 Seite 35 of 25 mm (50 mm in diameter). Hence the minimum sensor-roller distance is ~20 mm plus ~35 mm to be analysed over the lifetime of the roller. After the lab-studies in Part 1 [9] , two ultrasonic sensor configurations and a laser triangulation sensor were chosen for the field test: ■ Ultrasonic sensor: Baumer, UNCK 09U6914/ D1, distance 3-150 mm, resolution 0.3 mm, with beam nozzle ■ Ultrasonic sensor: Baumer, UNDK 09U6914/ D1, distance 3-150 mm, resolution 0.3 mm, with beam nozzle, angled shape ■ Laser sensor: Panasonic, HG-C1100, distance 65- 135 mm, accuracy 70 µm 2.2 Field test of sensors for HPGR A HPGR grinding cement clinker was chosen for real field tests, equipped with rolls with hexagon-shaped wear protection plates (Figure 1). The wear protection plates have a wrench width of 50 mm and ~5 mm gap. This structure leads to the minimum requirement to detect a lost hexagon. Ideally also a change in the hexagon’s shape, e.g. edge blunting should be detected. The sensors were mounted at the back side of the fixed roll on a rigid steel bar at a given location. Both ultrasonic sensors are basically identical and differ only by their mounting and case design. For this reason no further distinction is made between these two sensors. The output signal is designed as a voltage signal, which was recorded with a 9205 cDAQ module in conjunction with a 9174 cDAQ chassis from National Instruments. At these first tests the sensors were deliberately not covered to ensure the worst case scenario relating to dust and abrasive contamination. Over a period of ~1.5 hours eight measurements have been done to evaluate various situations and stages of contamination. The measurements are shown in the timeline in Figure 2 and chronologically numbered. The first measurement (#1) was done before any grinding, but after the roll’s engine had been started and reached its full speed. At this time the air, as well as the sensors were still dust-free. Hereby the basic functionality of the sensors could be checked and the data could be used as a reference for comparison with measured data during grinding operation. According to the plant operator, it takes about ten minutes until the grinding operation has stabilised. Hence, the first ten minutes from the milling-start on has been completely recorded (#2). Subsequently every ten minutes, intervals of two minutes where recorded (#3-7). Then the grinding operation was stopped (after ~75 min) and the last measurement (#8) was done when the air still contained significant amounts of dust. Aus Wissenschaft und Forschung 36 Tribologie + Schmierungstechnik · 66. Jahrgang · 4/ 5/ 2019 DOI 10.30419/ TuS-2019-0022 However, the surfaces of the grinding rolls experience severe wear, dominated by abrasive wear [5] and surface fatigue accompanied by material spalling or breakout due to excessive mechanical loads. Wear-resistant sleeves, hardfacings [6], rippled or studded profiles are protective measures to reduce wear, which is in a range of 0.1 - 10 g per ton of ground material [7] and which leads to material losses up to 50 mm in diameter. Ensuring continuous and safe operation without risks of sudden breakage as well as evaluating and monitoring the continuous wear behaviour are crucial to improve the performance and reduce maintenance costs [8]. 1.2 Wear sensors for abrasive applications A detailed review of sensors which could be used in such application is given in Part 1 of this work [9]. In the following passage a short summary will be presented: For wear situations to be expected in HPGR 3 types of sensors are applicable: i) inductive distance sensors. Those are very rigid and insensitive to non-conductive materials, i.e. adherent material and dust in the air will not alter the signal. Unfortunately, the large distances entail a huge measurement spot and impair the measurement accuracy, hence they will not be further discussed here. ii) Ultrasonic distance sensors use the air as transmission medium. They are relatively insensitive to dust, but may detect larger abrasives flying through the air. The measuring spot increases with increasing measurement distance. Focusing techniques, e.g. a beam nozzle, can be used to reduce the measuring spot size. This measurement technique is quite sensitive to tilted surfaces. iii) Laser triangulation sensors use a focused laser beam for measurement, hence they feature a small measurement spot. The diffuse reflection of the surface used for the distance calculation allows for very high tilting angles. A drawback is that they are sensitive to dust contamination. The aim of this work is to establish online wear sensors for harsh environments such as in HPGR operation. Three wear phenomena should be distinguished on the rollers: reduction in diameter due to continuous wear, outbreaks of wear protection material, and a crack through the roller. The roller should be monitored continuously during operation, i.e. the environmental conditions prevailing in the roller press must not alter or impair the measurement. 2 Experimental 2.1 Utilised distance sensors During HPGR operation, a movement of the roller through acentric loading and bearing clearance in the 10 mm range is expected, and a security distance of ~20 mm was added in case abrasive particles passing the measurement area. The structure depth of the roller’s surface was in the range of 10 mm. The wear a roller can experience before undergoing maintenance is in the range TuS_4_5_2019.qxp_T+S_2018 23.08.19 13: 15 Seite 36 2.3 Data post-processing Figure 3a shows representative data measured with one ultrasonic sensor. The offset of the data was removed for better representation. The dataset shows one revolution of a roll, which obviously runs acentric by ~10 mm. As expected the data includes some faulty measurements. A simple error detection and correction algorithm has been written for better data-visualisation. Most of the measuring errors of the ultrasonic sensor data are thought to be caused by the loss of the signal due Aus Wissenschaft und Forschung 37 Tribologie + Schmierungstechnik · 66. Jahrgang · 4/ 5/ 2019 DOI 10.30419/ TuS-2019-0022 Figure 1: Field tests at HPGR: a) back side of fixed roll with hexagon-shaped wear protection plates; b) mounted laser triangulation and two ultrasonic sensors on a rigid steel bar to measure the distance to the roll’s surface Figure 2: Timeline of measurements at the HPGR Figure 3: Data correction procedure: a) representative data recorded with an ultrasonic sensor containing faulty measurements; b) the faulty measurements replaced by estimated data a b a b TuS_4_5_2019.qxp_T+S_2018 23.08.19 13: 15 Seite 37 derations for possible signal post-processing and how they could be used in future. n fs ∙100 The failure rate FR = ------- was introduced to avoid n ts relying solely on a visual/ subjective evaluation of the data. It represents the percentage of faulty data within a data set, where n fs is the number of faulty data points and n ts is the total number of data points in a data set. It provides an objective comparison of the quality of two data sets. 3 Results 3.1 Results of the field test In Figure 4 representative data from the ultrasonicand the laser triangulation sensor can be seen (offset set to 0 Aus Wissenschaft und Forschung 38 Tribologie + Schmierungstechnik · 66. Jahrgang · 4/ 5/ 2019 DOI 10.30419/ TuS-2019-0022 to tilted surfaces. When the sensor loses the signal it returns its maximum distance (in our case ~80 mm from the original surface, Figure 3a). These measuring faults are easy to determine by choosing a threshold level and declaring all above data as faulty. For a small amount of data points the signal is not lost but falsified, probably by debris flying past. For this reason not a fixed single value was used as threshold, but the threshold was built up offline by fitting a polynomial function (green coloured line) in all data points smaller than a fixed value and applying this function with an offset (in our case 5 mm, red coloured line). After using the threshold to identify the faulty measuring points, they can be replaced by estimated data. One of the easiest ways to do so is to remove the faulty points and linearly interpolate the remaining points, as shown in Figure 3b. The discussion section 4.2 presents consi- Figure 4: Data of ~10 revolutions at a HPGR recorded at measurement #3: a) by ultrasonic sensor and b) by laser triangulation sensor a b Figure 5: Progress of failure rate at the measurements according to Figure 2 for better visualisation). Both were measured during operation at the HPGR (measurement #3), whereby each dataset shows ~10 revolutions. At Figure 4a the ultrasonic sensor signal losses are well visible as peaks leaving the expected surface structure range. For the laser sensor a few short-term faulty measurements, probably adhesion-related or due to flying debris, can be seen in Figure 4b as valleys (detected feature, e.g. abrasive particle, decreases the distance to the sensor). 3.2 Failure rate in field test through environmental influences Figure 5 shows the progress of the FR at the field measurements at the HPGR. All sensors deliver an acceptable low failure rate of <10 % for the whole test duration. Both ultrasonic sensors (red and blue lines) started at a failure rate of <2 %, which increased after opening of the material gate. As expected, dust and abrasive impair the measurement. Obviously the dust and particle contamination is less extreme in the application than simula- TuS_4_5_2019.qxp_T+S_2018 23.08.19 13: 15 Seite 38 ted in the laboratory experiment [9]. About 20 min (measurement #3) after opening the gate a steady state was reached, which was held even after closing the gate again (#8). The different steady state levels of the ultrasonic sensors can be explained by the different mounting places of the sensors: probably the straight sensor was placed on a more beneficial position. The FR of the triangulation sensor data is less than the FR of the other sensor data almost over the whole time. Actually, the entire time after opening the material gate the FR is below 0.1 % which is much lower that the equivalent FR of the ultrasonic data. An interesting fact is, that the FR drops after adding milling material to the process, which will be discussed in 4.1. 3.3 Surface representation by the different sensors An important characteristic is the ability to resolve features like cracks, welding seams, gaps between pads or the pads themselves, although it is not the goal of this work to resolve the details of one pad. Figure 6a shows a short detail of the laser triangulation data and as interpretation aid a representative image of the roll’s pad structure is shown in Figure 6c. The spikes marked with “A” display the gaps between the pads, so the resolution of the laser triangulation sensor even in the harsh HPGR environment is good enough to detect structures of <5 mm width. The minima (least distance to the sensor) marked with “B” can be interpreted as peaks of pads or least worn areas. As it can be seen in Figure 6c most of the pads are not uniformly flat-worn, some of them have an approximately triangular cross-section with a distinctive tip. The line “C” shows the almost flat, sloping plane after the peak. However, in the middle of these planes often a peak occurs (marked by “D”), although no equivalent real counterpart was found on the pads. Hence, it has to be regarded as measurement artefact. The discovery of the cause of these measurement errors will be part of future research. The quality of the ultrasonic sensor data (Figure 6b) is worse compared to the laser triangulation sensor data. It would be feasible to determine a general wear trend but it is not possible to reliably differentiate between two pads. For the detection of small objects like gaps, cracks, welding seams, etc. the measuring spot is too big. 3.4 Susceptibility of sensors to dust contamination Figure 7 shows the sensors directly after reopening the mill’s door after the test cycle, which consisted of ~75 min of operation. Obviously there was some dust deposition on the sensors and the mounting bar. But even if they were covered with dust, the sensors did not show any damage and also the function of the sensors remained intact. Figure 7b shows the front side of the laser triangulation sensor. At the corner between the sensor and the mounting bar dust was deposited, but not to such an extent that this would have impaired the measurement. The dust also adhered to the vertical smooth laser entry and exit surfaces. Even if the sensor gets covered, it has to be ensured that the vision panel remains dust-free for unaffected measurements in the future. Aus Wissenschaft und Forschung 39 Tribologie + Schmierungstechnik · 66. Jahrgang · 4/ 5/ 2019 DOI 10.30419/ TuS-2019-0022 Figure 6: Detailed view of measured data: a) data of laser triangulation sensor; b) data of ultrasonic sensor; c) representative image of the roll’s pad structure The marked positions represent: a gaps between two pads; B peaks of pads (least worn areas); C flat, sloping plane of a pad; D measurement artefact TuS_4_5_2019.qxp_T+S_2018 23.08.19 13: 15 Seite 39 and after opening the material gate were determined using FFT with Hann window [10]. Figure 8a shows the frequency spectra of the measured signals before the grinding action and Figure 8b when grinding was active. Both signals contain frequency components at about 0.3 Hz, which corresponds to the rolls revolution speed. A second distinct peak was found at ~20 Hz, showing the pad-structure. At ~40, 60, 80 and 100 Hz the harmonics and mirrored harmonics (aliasing effect [11]) of the pad structure can be seen. Just the frequency component at 50 Hz (Europe’s mains frequency) can be traced back to an electromagnetic interference, though the amplitude is the same at both spectra’s, thus it cannot be the reason for the FR-difference. Hence the reason of the improvement has to be in the measurability of the surface itself. It is assumed that the adhesion of dust to almost polished, reflecting pad surfaces improves the diffuse reflection of the laser dot, allowing for an easier detection of steep flanks such as those found at the pads’ edges on the roll. So in this case the dust contamination even has a beneficial effect on the measurement’s accuracy. 4.2 Possible future error correction and data processing methods The use of the measuring system in a HPGR during regular operation requires a robust, reliable and precise online data processing. It is important to be able to differentiate between faulty data and correct measured data, in particular if there are material break-outs or cracks to Aus Wissenschaft und Forschung 40 Tribologie + Schmierungstechnik · 66. Jahrgang · 4/ 5/ 2019 DOI 10.30419/ TuS-2019-0022 Also the ultrasonic sensor would need a specific measure for of dust protection. Demounting the beam nozzle shows that some dust was deposited in there, (which had no influence on the measurement). When thinking of a future long term implementation of the sensors in the plant, suitable methods have to be taken to ensure no dust contamination at the sensitive spots. Possible designs will be discussed in section 4.3. 4 Discussion Some general points arose when evaluating the field test results: ■ Why does the failure rate for the laser triangulation sensor decrease when the HPGR is in operation? ■ Which more advanced failure correction strategies can be applied compared to the simple one used here? ■ How can constant cleaning of the sensors be accomplished, ensuring a long-term implementation of the sensors? ■ Which sensors are best to be used in these harsh environments for reliable maintenance assistance? 4.1 Decrease of failure rate at laser triangulation sensor when HPGR is in operation The failure rate of the laser triangulation sensor was in the range of 1.2 % with no abrasive (Figure 5). From this already low level it dropped further to almost 0 % when the HPGR was in operation. To rule out the possibility of random electromagnetic interference in the first few minutes of the laser triangulation sensor signal, the frequency components of the measured signals before Figure 7: Dust-coated sensors after use in the mill: a) ultrasonic sensors from the backside; b) laser triangulation sensor from the front side Figure 8: FFT transformed laser triangulation sensor data: a) measurement #1 without material grinding; b) measurement #2 with material grinding a a b a b TuS_4_5_2019.qxp_T+S_2018 23.08.19 13: 15 Seite 40 be expected during operation, which may appear quite similar to faulty measurements. There to detect incorrect measurements the definition of boundary conditions is feasible. Thereto the fact could be used that wear usually takes place very slowly. If the angular position of the roll is known, the same roll position can be detected again with each revolution. This makes it very easy to distinguish between: constant, very slowly changing geometry, random measurement errors due to debris flying past, and short-term adhesions. Just the occurrence of break-outs or cracks can cause an abrupt change of the measurement data during one roll revolution. However, these changes remain permanently in the measurement data, even if they are filled up to a certain extent with milled material. Referring to this a further condition can be used: The roll diameter can only increase if the temperature rises or a substantial crack occurs. Both mechanisms can be clearly identified. When a crack occurs, the diameter increases instantly over almost the entire circumference of the roll. In contrast to material build-up which only leads to a local change in distance. Diameter variation due to temperature changes never occur abruptly and can be evaluated by additional temperature monitoring. In case that it is necessary to assess areas that cannot be monitored over a longer period, e.g. for geometric reasons, or in a test measurement like presented here, special algorithms can be used to replace the missing data with estimated values. Next to the linear approximation e.g. the recursive least squares method with exponentially decreasing memory [12] would be suitable in this case. 4.3 Constant cleaning of sensors Already after the relatively short implementation of the sensors in the HPGR (~75 min of operation) significant dust deposits were built at critical positions. For a longterm operation suitable measures have to be taken to avoid or constantly remove dust from the critical sensor areas, i.e. the laser outlet and viewing window of the laser triangulation sensor and the beam nozzle of the ultrasonic sensor. For the laser sensor an encapuslation in a transparent housing will avoid dust deposition, but the problem can also occur on the surface of the protection box. An encapsulation of the ultrasonic sensor is not possible, as the ultrasonic beam needs free field to the measured object. Hence a shielding out of non-solid material is necessary, e.g. a constant flow of clean air. As compressed air is usually available at industrial plants, its implementation should be possible. Two problems have to be kept in mind when cleaning with a constant flow of compressed air. Firstly, the production and continous use of compressed air entails additional costs and power consumption. Hence, the airflow should be minimised to the absolute necessary amount. Secondly, for the ultrasonic sensor, strong airflows can lead to measurement errors, as they influence the sound propagation. We made simple experiments to determine if an influence is detectable for our sensor. No measurement errors occured even under the influence of strong air flows of a compressed air gun operated with 65000 Pa that is pointed directly to the sensor. 4.4 Applicability of the tested sensors for maintenance assistance in harsh environments In the harsh environment of heavy industries the wear of components is a crucial factor influencing the maintenance intervals. Often key components are changed prematurely because a sudden breakdown cannot be risked. In such cases the monitoring of the current status of these components decrease maintenance costs and resources [13], [14]. In such applications the environment is mostly very harsh [e.g. [15], [16]], e.g. with high abrasive load, high temperatures, etc., which makes the application of sensors challenging. The application of local wear protection, often with brittle materials which can break-out [17], [18] also requires a localised wear monitoring. When moving components need to be monitored also large distances may be necessary for the safe placement of sensors. These factors generally impede measurements of high accuracy which are implemented e.g. in laboratory tribotests for wear monitoring [e.g. [19], [20]]. Hence, the necessary resolution required for maintenance interaction has to be considered. In our case the detection of abrasive wear loss, the loss of single wear protection plates and large scale cracking of a roll should be detected. In terms of distance sensors these requirements are not very high, but they have to be fulfilled during the plant operation and maintained for years. Although the ultrasonic sensors were able to measure the distance to the surface correctly (i.e. wear loss can be detected), their measurement spot was too large to detect the surface structure. The measurement spot at the required sensor distance is in the range of 10-12 mm, which means that the loss of a complete wear protection pad can be detected. Concerning the minimum requirements, we think that the ultrasonic sensor can fulfil them at the necessary accuracy. Most of the details of the surface structure were measured with the laser triangulation sensor. Overall, it is possible with the used laser sensor to detect objects on the roll from a lateral width of at least 5 mm (e.g.: gaps in the present case), i.e. all requirements are fulfilled. A further improvement would be possible with a higher sample rate of the sensor, as in this case also details of Aus Wissenschaft und Forschung 41 Tribologie + Schmierungstechnik · 66. Jahrgang · 4/ 5/ 2019 DOI 10.30419/ TuS-2019-0022 TuS_4_5_2019.qxp_T+S_2018 23.08.19 13: 15 Seite 41 consumption by avoiding prematurely component exchange can be achieved by online wear measurement. Acknowledgements This work was funded by the Austrian COMET Programme (Project K2 XTribology, Grant No. 849109) and supported by the Province of Niederösterreich (Project “Digi-Pro”, WST3-F-5030642/ 004-2018) and has been carried out within the “Austrian Center of Competence for Tribology” (AC2T research GmbH). References [1] F. W. Locher, Zement: Grundlagen der Herstellung und Verwendung, Düsseldorf: Bau+Technik Verlag, 2000. [2] R. Dunne, “HPGR - the journey from soft to competent and abrasive,” in International Autogenous and Semiautogenous Grinding Technology Conference, Vancouver, Canada, 2006. [3] N. A. Aydogan, L. Ergün, H. Benzer, “High pressure grinding rolls,” Minerals Engineering 19, pp. 130-139, 2006. [4] J. A. Drozdiak, B. Klein, S. Nadolski, A. Bamber, “A pilot-scale examination of a high pressure grinding roll/ stirred mill comminution circuit,” in International Autogenous and Semiautogenous Grinding Technology Conference, Vancouver, Canada, 2011. [5] H. Rojacz, G. Mozdzen, F. Weigel, M. Varga, “Microstructural changes and strain hardening effects in abrasive contacts at different relative velocities and temperatures,” Materials Characterization 118, pp. 370-381, 2016. [6] M. Varga, “High temperature abrasive wear of metallic materials,” Wear 376-377, pp. 443-451, 2017. [7] T. Hanstein, Beitrag zur Erhöhung der Standzeiten der Arbeitsorgane von Gutbett-Walzenmühlen, Dissertation TU Bergakademie Freiberg, 2001. [8] L. Widder, S. Leroch, M. Kirchgaßner, M. Varga, „Finite Elemente-Simulation als Werkzeug für ein spannungsgünstiges Design von Hochdruck-Rollenpressen in der Zementindustrie,“ Berg- und Hüttenmännische Monatshefte 163/ 5, pp. 181-187, 2018. [9] M. Varga, R. Grundtner, A. Maurer, M. Kirchgaßner, “Online wear measurement in harsh environment. Part 1: Possible measurement strategies,” Tribologie und Schmierungstechnik, 66. Jahrgang, pp. 36-42 4/ 5/ 2019. [10] M. Vorländer, Digitale Signalverarbeitung in der Messtechnik, Berlin Heidelberg: Springer, 2018. [11] A. Meroth and P. Sora, „Sensortechnik,“ in Sensornetzwerke in Theorie und Praxis, Wiesbaden, Springer Fachmedien, 2018, pp. 219-341. [12 R. Isermann and M. Münchhof, Identification of Dynamic Systems, Berlin Heidelberg: Springer, 2011. [13] M. Varga, K. Adam, R. Wimberger, E. Badisch, “Cost efficient tribological systems in steel production based on Aus Wissenschaft und Forschung 42 Tribologie + Schmierungstechnik · 66. Jahrgang · 4/ 5/ 2019 DOI 10.30419/ TuS-2019-0022 the pad’s structure and gaps would be resolved. But, concerning the requirements this is an overfulfilment and not a requisite for wear monitoring of an HPGR. Permanent implementation of a suitable sensor in a plant like the HPGR will contribute to a detailed prediction of the necessary maintenance actions and the remaining lifetime of the component. In the present case a critical threshold of lost wear protection plates can be defined, making a repair necessary when the threshold is reached. To monitor the roll’s whole surface the installation of a higher number of sensors or one sensor in combination with a linear positioning unit will be needed. For the lifetime of a roller the overall wear loss is crucial, i.e. the decrease in the rolls diameter [7]. For continuous production processes it is expected that this wear loss will be relatively constant. Hence, from a longterm recording the average wear rate can be accurately determined. From the knowledge of the total wear volume available, a remaining lifetime of the component can be predicted easily. In the case of changing operation conditions also a number of different wear rates can be taken into account for a precise determination of the remaining lifetime. E.g. a lab-2-field approach can be utilised to determine the different occurring wear rates [21]. The knowledge of the remaining lifetime is especially important for components like the large HPGR rolls, as they usually have several months of delivery time. Hence, the implementation of wear sensors can make maintenance much more efficient and better schedulable. 5 Conclusions In this work the online measurement of wear in the context of a high pressure grinding roll (HPGR) has been studied. Following major findings can be drawn from the field measurements: ■ The structure of the HPGR was resolved in great detail with the laser triangulation sensor. The hexagonal wear pads and the gaps between them were measured. Even the topography of a single wear pad could be resolved with this sensor. ■ The overall wear loss can be detected with the ultrasonic sensors. Due to the large measurement spot diameter they were not able to resolve the surface structure. ■ Dust contamination was the most detrimental factor. Both, ultrasonic and laser triangulation sensors are sensitive to it. As suitable measure flooding of the neuralgic positions with compressed air is proposed. Long-term implementation of online wear measurement in industrial components will allow for accurate lifetime prediction and schedulable maintenance intervals. An increase of plant profitability and decreased resource TuS_4_5_2019.qxp_T+S_2018 23.08.19 13: 15 Seite 42 life cycle optimisation,” in proceedings 5 th World tribology Congress, Turin, IT, pp. 1921-1924, 2013. [14] M. Varga, M. Haas, C. Schneidhofer, K. Adam, “Wear intensity evaluation in conveying systems - An acoustic emission and vibration measurement approach,” Tribology International, https: / / doi.org/ 10.1016/ j.triboint.2019. 01.008, in press, 2019. [15] M. Varga, L. Widder, M. Griesinger, K. Adam, E. Badisch, “Wear progress and mechanisms in high temperature sieves,” Engineering Failure Analysis 61, pp. 46-53, 2016. [16] H. Torres, M. Varga, K. Adam, E. Badisch, “Wear phenomena in high temperature sheet shearing blades,” Wear 306, pp. 10-16, 2013. [17] H. Rojacz, M. Varga, H. Kerber, H. Winkelmann, “Processing and wear of cast MMCs with cemented carbide scrap,” Journal of Materials Processing Technology 214, pp. 1285-1292, 2014. [18] M. Varga, A. M. F. Azhaarudeen, K. Adam, E. Badisch, “Influence of load and temperature on abrasion of carbidic cast steel and complex alloyed hardfacing,” Key Engineering Materials 674, pp. 313-318, 2016. [19] S. B. Glavatskih, Ö. Uusitalo, D. J. Spohn, “Simultaneous monitoring of oil film thickness and temperature in fluid film bearings,” Tribology International 34, pp. 853-857, 2001. [20] G. Garcia-Atance Fatjo, E. H. Smith, I. Sherrington, “Mapping lubricating film thickness, film extent and ring twist for the compression-ring in a firing internal combustion engine,” Tribology International 70, pp. 112-118, 2014. [21] P. O. Bedolla, G. Vorlaufer, C. Rechberger, D. Bianchi, S. J. Eder, R. Polak, A. Pauschitz, “Combined experimental and numerical simulation of abrasive wear and its application to a tillage machine component,” Tribology International 127, pp. 122-128, 2018. Aus Wissenschaft und Forschung 43 Tribologie + Schmierungstechnik · 66. Jahrgang · 4/ 5/ 2019 DOI 10.30419/ TuS-2019-0022 TuS_4_5_2019.qxp_T+S_2018 23.08.19 13: 15 Seite 43