Laser-Structured Thin Film Sensor Technology for Sliding Bearings in Internal Combustion Engines

Global issues such as climate change, environmental pollution, and conservation of resources require manufacturers of ICE to achieve substantially reduced emissions of greenhouse gases (GHG) and pollutants as well as increased engine efficiency and durability [1–3]. Condition monitoring (CM) and predictive maintenance (PdM) approaches for ICEs are key tools for increasing engine durability and saving resources by utilizing more of the useful lifetime of engines and components thereof while avoiding critical operating conditions caused by wear and failure [4]. In the recent past, the extensive instrumentation of large ICEs and their subsystems as well as the availability of advanced methods for data analytics from the field of machine learning has led to a large variety of CM and PdM concepts for entire engines [5–9] as well as for individual components such as fuel injection valves [10–12], cylinder liners [13–15], torsional vibration dampers [9,16], elastic couplings [16,17] and sliding bearings [18–20].

Sliding bearings play a key role in ICEs to support shafts such as crankshafts or camshafts [21]. Recent developments to improve ICEs with regard to fuel consumption as well as GHG and pollutant emissions through low viscosity oils (friction reduction) and start-stop technology (avoidance of idling) pose new challenges for these bearings and intensify the need for corresponding CM and PdM concepts [20,22]. These concepts require appropriate measurement technology capable of acquiring informative parameters that reflect the current condition of the bearings while withstanding the high temperatures and mechanical loads encountered inside the engine and accommodating spatial constraints [20,23,24].

In the existing literature, several measurement parameters which can reflect bearing condition have been discussed, such as oil contaminants [25,26], vibration and acoustic emission [27–29], bearing shell temperature [30], oil film temperature [31], pressure [32–34] and thickness [31,35] as well as metal-to-metal contact [28,36].

While all measurement parameters focus on the interactions between the rotating shaft, lubricant oil, and inner surface of the bearing shell, each of them comes with different characteristics with regard to information content, response time, sensor size, as well as overall measurement chain complexity and cost.

In order to obtain significant information from specific, individual bearings and to ensure high signal information content as well as short signal response time (for capturing highly transient phenomena in real-time), it is desirable to minimize the distance between the bearing area of interest and the sensor position. This in turn leads to a challenge in instrumentation, since the particular areas of interest (highly loaded parts of the bearing running surface and/or the lubrication gap) are difficult to access with conventional instruments which must not influence bearing functionality [19].

For these reasons, thin film sensors were identified as an advanced instrumentation technology that can be highly integrated into the bearing (and its production process) and allows to obtain information quasi-directly at the bearing running layer while the mechanical and hydrodynamic properties of the bearing remain unaffected [18,37,38]. Such sensors are characterized by a thin layer of sensor material and a corresponding sensor structure that is positioned closely below the bearing running layer [33,34]. Conceivable parameters for bearing CM are restricted by the measurement principles which can be realized with help of a sensor layer. Such principles build upon measuring changes of physical properties of the sensor material caused by a change in the measured parameter (e.g., electrical resistance change due to temperature).

Despite the promising approach, a comprehensive literature study indicates that only few researchers have made attempts to employ thin film sensors on sliding bearings. Mihara et al. [39,40] have used sputtered thin film strain sensor structures made of alloy to measure oil film pressure on an engine main bearing, on a conrod big end bearing and on a piston pin boss for research purposes. Kataoka et al. [33] used a similar sensor setup on the main bearing of a passenger car diesel engine to validate calculated oil film pressure traces with help of corresponding experimental results. Such validation of simulation results was also performed by Miura et al. [34], employing the abovementioned thin film sensor technology on a piston pin boss in test rig investigations. With regard to temperature measurement, Lichtenwalner et al. [41] have developed a corresponding thin film sensor concept, which, however, has not been applied to sliding bearings. Previous research work on thin film sensors for sliding bearings by the Large Engines Competence Center (LEC GmbH) and its research partner consortium was presented by Breiteneder et al. [38] and covers early concepts for sensor layer and sensor structure manufacturing as well as for sensor layouts for strain measurement.

The objective of the present publication is to introduce the latest developments for an advanced thin film sensor concept for sliding bearings in ICEs that allows to monitor strain and temperature in relevant areas of interest. These parameters were chosen because they are considered both informative for bearing CM and PdM purposes as well as achievable with help of the thin film sensor technology. Promising sensor layouts and positions targeted for use in CM and PdM applications are elaborated in detail. Developments and challenges in implementing the sensor technology concept—in particular with regard to the process of manufacturing the sensor as well as the wire contacting—are outlined in depth. The paper concludes by presenting measurement results obtained with the investigated sensor technology at lab scale as well as an outlook toward implementing the instrumented bearings in ICEs.

This section deals with considerations for designing thin film sensors for bearing CM and PdM purposes. Such considerations include the choice of material and measurement parameters as well as position, size, and detailed structure of the related sensors. In the present study, the focus is placed on sliding bearings that are typically used as crankshaft main bearings and conrod big end bearings in heavy-duty ICEs and feature an inner diameter in the range of approximately 75–105 mm.

With regard to sensor material, the following considerations were made: A material must be chosen that can be applied with help of a suitable manufacturing process to generate a corresponding thin film sensor layer. This material must be able to bond with the surrounding insulation layers (cf. Sec. 3.1), which have to provide the best possible insulation for the sensor layer between them. Furthermore, the electrical properties of the sensor layer material must be suitable for the selected measurement principle. Based on previous work at LEC [38], nickel–chromium (NiCr) was selected as the sensor layer, since it bonds with the material chosen for the insulation layers (aluminum oxide, Al₂O₃) and can be applied to the bearing shell with help of a corresponding sputtering process and machined by a laser process, cf. Sec. 3.2. In its chosen composition (80% nickel, 20% chrome), the material has characteristics which make it particularly suitable for strain measurement, since its resistance change due to temperature is low (albeit linear) and its thermal expansion coefficient matches well with that of steel, which serves as the bearing shell base material (cf. Sec. 3.1).

With these qualities, strain measurement is considered as the first measurement parameter to be obtained for bearing CM and PdM purposes. Though beyond the scope of this study, strain could be correlated with the mechanical load that acts on the bearing. Continuous recording of the amplitude of load cycles as well as their total number could serve to detect anomalies during engine operation (e.g., by comparing with reference bearing load conditions at comparable engine operating settings) and to evaluate bearings with regard to component fatigue (e.g., by running a related model in parallel that predicts bearing remaining useful life based on the real component's history). While it was already outlined that the resistance change of NiCr due to temperature is low, the effect can still be employed for measuring temperature. The temperature at the running surface provides significant information about the state of the bearing, since unwanted metal-to-metal contacts during engine operation inevitably lead to an increase in temperature at the bearing surface [29]. Temperature is therefore the second measurement parameter that is considered for implementation by means of thin film sensor technology in this publication.

To obtain a foundation for the design of sensors with regard to their position, size and alignment, preliminary investigations were carried out to understand which loads are to be expected in sliding bearings in a heavy-duty ICE and how the bearings deform as a result of these loads. Thereby, the focus was placed specifically on crankshaft main bearings. The preliminary investigations were based on the available literature on the subject of sliding bearings, simulation results on the pressure distribution of the oil film and on finite element (FE) simulation models for the corresponding deformation of the bearings.

The basic distribution of pressure in the oil film of a sliding bearing that is loaded by a force with a specified direction is described in the existing literature [21,42,43]. A pressure peak builds up on the side facing away from the load, with the maximum of the pressure distribution at an offset of approximately 15 deg from the bearing apex opposite of the load in the direction of shaft rotation. The lubrication gap is also smallest in this area. In an ICE, the direction of the force changes since the gas pressure from the combustion chamber is transmitted radially into the crankshaft bearing via the crankshaft drive. During a working cycle, the main bearings are loaded by the combustions from all individual cylinders, but it is expected that the load from the directly adjacent cylinders will be the greatest. It is therefore assumed here that the maximum pressure in a main bearing occurs at an offset of approximately 20–25 deg from the apex opposite of the combustion chamber, whereby this is made up of the addition of the load direction due to the maximum cylinder pressure (several degrees crank angle after top dead center) and the oil film pressure peak offset due to the hydrodynamic processes.

At full load, a radial force of 65 kN acting on the bearing can be expected in a typical heavy-duty engine. According to previous elasto-hydrodynamic (EHD) simulations of the lubrication gap at Miba Gleitlager Austria GmbH, pressure peaks with a maximum of up to 3000 bar may be expected. This value was used as a boundary condition for the FE simulations with Abaqus 2017 in this study to determine bearing deformation. The geometry used in the FE simulations originates from a heavy-duty in-line 6-cylinder diesel engine with a displacement of 12.4 liters. It consists of a three-dimensional model of the lower (loaded) bearing shell and the corresponding bearing cap that is mounted underneath the crankshaft and accommodates the shell (Fig. 1).

The shell is meshed circumferentially with C3D20R hex elements of 0.5 mm and the cap with C3D4 elements of max. 5 mm. A mesh independence study has revealed that the elements are sufficiently small. The bearing shell is not exactly circular when unmounted, but has a slightly larger diameter at its two ends than at the center of the shell. In addition, the outer diameter of the shell is slightly larger than the diameter of the supporting bearing cap. The shell is therefore already deformed during installation in the bearing cap, resulting in a small protrusion of the shell over the cap. When the bearing cap is tightened to the crankcase, this protrusion diminishes and a large preload is created, preventing the shell from twisting during operation. The resulting contact pressure between the shell and the cap is in the range of 16 N/mm².

After simulating the assembly process, the two ends of the bearing shell are no longer allowed to move vertically, while the separating surfaces of the bearing cap are now fixed in vertical and horizontal directions, and a cosine-shaped pressure distribution with a maximum of 3000 bar is applied to the inside of the bearing shell (Fig. 2). The cosine distribution of the pressure represents a simplified yet similar pressure distribution compared to that described in the literature. The maximum pressure is either applied at the lowest point of the bearing shell (0 deg) or at an offset of −30 deg since it is expected that the deformation behavior of the bearing can be well characterized using these two positions.

The result of the simulation was evaluated with respect to the deformations on the inner side of the bearing shell, since it can be assumed that a thin film sensor applied to this very surface by sputtering will undergo the same deformations as the bearing itself in this area. In Fig. 3, the relative distances of the respective neighboring nodes on the inner surface of the computational mesh are plotted along the symmetry plane of the bearing as well as at the edge of the shell (cf. “center” and “edge” in Fig. 2). These distances change from the initial state (undeformed bearing, 100%) due to the deformation of the shell during the installation into the bearing cap and to the engine as well as during the application of the pressure load in the lubrication gap. Since the electrical resistance of the sensor is expected to change in proportion to the elongation or compression experienced by the individual wires of the sensor, the relative change in node spacing can provide an estimation of the sensor signal.

The dark gray curve in the diagrams (Fig. 3) shows the node spacing in the undeformed state of the bearing shell. The dashed lines show the condition after the bearing shell has been installed in the bearing cap, and it can be seen that the node spacing decreases as a result of the installation and that the most significant change occurs in the center of the shell. With the pressure load applied, it can be seen that the compression from the assembly process is now reduced on the inner shell surface due to the pressure load, but the reduction in compression is much more pronounced in the areas around +60 deg and −60 deg than at 0 deg. Since the pressure distribution is also cosine in the axial direction, the change compared to state after the installation process is greater in the center than at the edges of the shell. This reduction of the preload can be explained by the elongation of the bearing cap and the shell due to the applied pressure load. Since the deformation of the crankcase has not been considered, the reduction in preload is somewhat overestimated in this simulation. The “dent” at the 0 deg position indicates a local surface elongation due to the pressure load distribution.

If the maximum of the pressure load is assumed to be on the −30 deg position, basically same trends result as with the 0 deg position. If the deformation effect of the pressure on the bearing shell is to be measured with a sensor in which the main measuring direction is in the circumferential direction, it appears advantageous to use the installed (and deformed) shell as the baseline for an undeformed sensor to avoid using the available measurement range partly for the installation process.

In addition to the comprehensive model outlined above, two simpler models were generated to better understand the results. In the first of these simplified models, the shell and cap were modeled as a single combined part. Thereby, relative motion between the shell and the cap is prevented, so that the impact of global deformation of the cap on the deformation of inner surface of the shell due to the pressure load can be investigated. This model also lacks the preload created by the installation of the shell. In the second simplified model, the shell and the cap were modeled as separate parts, but the profile of the shell was considered perfectly circular and with the same diameter as the bearing cap, so that no deformation occurs during assembly. In addition, the bearing cap was modeled as completely rigid so that the effect of local squeezing of the shell due to the pressure in the lubrication gap could be analyzed.

The results of the simulations with the simplified models provide further qualitative information about the composition of the overall deformation and are illustrated based on the case with the pressure maximum at the 0 deg position. The simulation with the rigid bearing cap (Fig. 4, left) shows that the pressure load leads to a local elongation of the inner bearing surface. Because the preload is missing in this simulation, most areas of the shell experience elongation when the pressure load is applied. According to the simulation with the combined shell-and-cap-component (Fig. 4, right), the global deformation of the bearing cap under the pressure load leads to significantly stronger elongations of the bearing shell compared to the simulation with a rigid cap, yet these elongations occur only at the outer regions and not directly at the position of the pressure maximum at the center of the shell. When combined, the three partial deformations from the installation, the squeezing and the global deformation approximately yield the overall result from the comprehensive simulation.

As a result of these investigations, it can be stated that the installation-induced compression on the inner surface of the shell decreases when a compressive load occurs, but to a different extent in different areas of the shell. While for large signal output, large sensor areas would be beneficial, such local deformation phenomena must be captured with sufficiently small sensors. Therefore, it appears purposeful to place several smaller sensors along the circumferential direction and thus deduce the load on the bearing shell from the combination of the individual sensor signals. In addition, the sensors should be placed in the center of the shell in axial direction for two reasons.

First, wear can occur at the edges of the bearing during engine run-in, which could destroy a sensor in this area. Second, manufacturing the insulation and sensor layers has revealed that electrical insulation in the edge areas is comparatively poor, cf. Sec. 3.3.

Based on the considerations regarding the sensor position and size, a deformation sensor layout was designed. It consists of three individual strain sensors which are positioned at 30 deg intervals as illustrated in Fig. 5. Each sensor structure is a full-bridge strain gauge whose measuring circuits cover area of 16 × 16 mm (cf. Fig. 6(a)). The meander-shaped structures of these circuits feature two characteristic directions that enclose an angle of 90 deg, with two diagonally opposite circuits aligned identically in each case. This layout considers that the most relevant deformation of the bearing is due to bending around the shaft axis and that the largest deformations appear in the area where the highest oil pressures appear, cf. Sec. 2.2. The full-bridge configuration makes the sensor insensitive to temperature influence and the sensor area is designed for maximizing the deformation sensitivity of the sensor grid while avoiding areas of low insulation resistance at the bearing's edge regions.

The targeted sensor grid resistance is 1000 Ω. This target value is a compromise that considers the advantages of small grid resistance (low thermal noise and comparatively short sensor grids with a correspondingly small laser manufacturing time) as well as the advantages of large grid resistance (comparatively small current flow through the sensor circuits with correspondingly small self-heating of the grid and relatively small impact of passive resistances such as contact resistance and lead resistance between the contact pads and the bridge network). The contact pads are located in an area with a larger bearing shell diameter to provide space for wire contacting. A conically shaped transition area connects the two cylindrical surfaces on the inside of the bearing shell (cf. Fig. 5).

For temperature measurement, a separate sensor layout was designed in which the two outer strain sensors were replaced by sensor structures for four-wire resistance temperature detection (cf. Fig. 6(b)). The characteristic direction of the meanders of the measurement circuit was chosen to avoid impact of the expected mechanical deformation of the sensor structures due to bending around the shaft axis. As with the strain sensors, the targeted grid resistance is 1000 Ω.

For investigations carried out in parallel to sensor layout development (e.g., manufacturing processes for sensor layer, sensor structure and wire contacting), a sensor design from preliminary research work, which comprises a three-element delta rosette strain sensor structure (Fig. 7) was used and can therefore be found in several illustrations below.

This section introduces the manufacturing processes of the thin film sensor layers and the sensor structures and presents various challenges faced, in particular with regard to manufacturing processes as well as wire contacting.

The intended design of the bearing shell concerning the different bearing layers is illustrated in Fig. 8. Building on the steel back plate and a bronze layer, a thin film “sandwich” of two insulation layers and the sensor layer is applied. On top of that, a final layer acts as running layer. The three thin film layers relevant to the sensor are manufactured with help of sputtering, a physical vapor deposition process for coating of substrates in a vacuum [38]. Thereby, the material to be sputtered is brought into the vacuum in the form of a “target.” An argon plasma is ignited in the process chamber, with the target connected as the anode of this discharge. The resulting intense bombardment of the target surface with argon ions atomizes the target material and condenses it onto the substrates (i.e., the bearing shells to be coated) in the chamber. With this process, thin film layers with high adhesion can be produced at comparatively low temperatures.

In a step-by-step process, the first insulation layer and the sensor layer are manufactured. Thereafter, the coating process is inter-rupted and the sensor structures are manufactured (cf. Sec. 3.2). Subsequently, the remaining bearing layers are applied. For various topics investigated in this paper, only the first insulation layer and the sensor layer were manufactured, whereby the sensor structures are visible in the corresponding illustrations.

Ultrashort pulsed laser beams are capable to selectively ablate the NiCr film without damaging the underlaying insulation layer. In this work, a laser source of type Coherent Monaco is used, which delivers laser pulses with < 350 fs pulse duration at a wavelength of 1035 nm. The laser source is operated at a pulse repetition rate of 1 MHz which allows, in combination with a fast galvanometer scanning system, to achieve scanning speeds up to 1 m/s. The laser beam is focused by a f-theta scanning lens with a focal length of 160 mm onto the surface of the bearing, as shown in Fig. 9. In this configuration the laser focus has a diameter of 35 μm. A motorized focus shifter is synchronously controlled with the galvanometer scanner so that the beam focus follows the curved surface of the bearing. Additionally, an intensity modulator (integrated in the laser source) is dynamically controlled to compensate for effects induced by the varying angle of incidence of the laser beam during a scan process. On a tilted surface, e.g., on the transition slope between the two cylindrical surfaces, the circular laser focus is projected to an elliptical intensity distribution on the surface. Here, the process controller slightly increases the beam intensity as a function of angle of incidence to keep the peak intensity constant which ensures homogeneous ablation results.

The laser structuring of Wheatstone resistor bridges requires balancing of the resistor grids for high accuracy measurements. The bridge balance is affected by various factors such as film inhomogeneities, positioning errors of the laser beam, surface roughness, and pretensioning. Bridge balancing can be performed by a laser trim process (see Sec. 3.4). Since a bearing shell is assembled in the bearing cap with a considerable pretensioning, bridge balancing of relaxed shells may result in unbalanced bridges after assembly. Therefore, a tensioning device is integrated into the machining setup which allows to clamp a bearing shell in a deformation state similar to the final assembly in an engine.

A folding mirror deflects the laser beam into the bearing shell, as shown in Fig. 10. The mirror is mounted on a motorized rotation stage and moved during the laser process so that the laser beam can reach the entire inner bearing surface.

Figure 11 shows a detailed view of a laser processed isolation line. The laser beam is scanned in paths parallel to the contour of the sensor layout. The typical line thickness is 100 μm to ensure sufficient isolation between the sensor structure and the surrounding area. The process duration of a sensor system as shown in Fig. 5 is typically in the range of less than 10 min with potential to further optimization, e.g., by reducing the line width, introducing beam shaping techniques for increased pulse ablation rates, or optimizing the layout design with larger corner radii to process at higher scanning speeds.

To investigate the basic functionality of the first insulation layer, the sensor structure manufacturing process outlined in Sec. 3.2 was used to manufacture a bearing shell with a “checkerboard-patterned” sensor layer as illustrated in Fig. 12. Subsequently, the insulation resistance was determined for each of the individual squares with help of a digital multimeter that applies a constant current $I$ of 1 μA (at a maximum Voltage $V$ of 10 V) between the bearing back plate and sensor layer square. By measuring $V$⁠, the insulation resistance $R$ was calculated as $R=V/I$⁠. It was found that there is a trend that the insulation resistance is lower near the bearing shell edges, in particular in tangential direction, cf. Fig. 13. Based on the current state of knowledge, the changes in insulation resistance are likely due to the thin film layer manufacturing process not being capable of distributing the insulation layer uniformly throughout the entire relevant bearing shell geometry. Consequently, the pronounced decrease of insulation resistance in tangential direction was considered for the positioning of sensor structures on the bearing shell.

To study the functionality of the first insulation layer after manufacturing of an actual sensor layout, a similar test was carried out by applying a voltage of 25 V between each of the sensor grids and the bearing back plate. It revealed that proper insulation of the sensor grids from the bearing back plate is challenging. It seems likely that small imperfections of the insulation layer result in electric arcs through this layer when the voltage is applied. The sensor layer is then evaporated around such a point defect until current flow stops, which results in a “healing” of the defect. While all low-resistance defects can be removed from a sensor structure through this voltage procedure, the pinholes in the insulation layer are not closed and it is yet unknown if the remaining crates may have a negative impact on the sensor performance on the long term or after deposition of additional layers.

It was found that such damage in many cases appears in the conically shaped area of the bearing shell, cf. Fig. 14. Detailed investigations are currently being carried out to understand the root causes of this behavior. A preliminary investigation suggests that the original surface quality of the conical area (before the thin film layers are applied) area is inadequate and/or the proper application of the thin film insulation layer by means of the sputtering process is challenging due to the conical surface geometry.

A bridge balancing procedure was employed to reduce the electrical asymmetry of the sensor bridge down to the operating range of corresponding signal conditioning circuits. Signal conditioning circuits with high amplification factors may saturate if the bridge symmetry is not in the specified range and the useful measurement range can be reduced by a large bridge offset. Bridge asymmetry is a result of manufacturing tolerances, and the bridge can be balanced by adjusting the values of single resistors in the resistor network. A laser trim process allows for trimming the active sensor grids so that no additional trimming areas are required, and no passive trim resistors are included into the sensor network. For that purpose, the sensor circuits of the bearing were connected to a bridge conditioning circuit with the help of a detachable contact system while the bearing is still mounted in its original position for the laser machining process, cf. Fig. 15.

A trim feedback loop implemented in the machine controller iteratively measures the bridge values and runs a laser trim step which removes small amounts of material at the edge of the sensor structure. The trim process terminates when a desired stop criterion is reached, e.g., that the bridge offset is within an acceptable range. In the experiments in this study, the bridge offset was reduced by 2 to 3 orders of magnitude, i.e., initial offsets of up to several 100 mV/V were trimmed down to a remaining bridge offset of less than 1 mV/V. However, strong asymmetry of the sensor bridges has been observed which is attributed to manufacturing grooves on the surface of the bearings. This kind of systematic effects can be adjusted by the sensor layout, as discussed in Sec. 3.5.

Electrical testing of the sensor grid with the setup described in Sec. 3.4 revealed that the sensor grid resistance strongly depends on the orientation of the meanders since the turning grooves (Fig. 16) induce a directional anisotropy of the sensor layer sheet resistance. Since the same width of the grid lane is used for all four quarters of the sensor (uniform layout), this leads to a grid resistance that is approximately twice as large for grids perpendicular to the grooves compared to that of grids parallel to the grooves.

Based on the implementation of the measured sensor layer sheet resistance in both characteristic directions in a sensor grid simulation software, a correspondingly adapted nonuniform sensor layout was developed which achieves the same resistance in all individual grids through adoption of wider grid lanes in the direction perpendicular to the grooves and narrower grid lanes in the direction parallel to the grooves (cf. Fig. 17).

Based on the current state of knowledge, both sensor grid trimming process (cf. Sec. 3.4) and sensor layer resistance anisotropy countermeasures require to be managed on an individual bearing shell basis, which poses a significant challenge to manufacturing thin film instrumented bearing shells with narrow tolerances concerning the sensor specification.

The concept of contacting the sensor elements is based on a process of establishing solder joints between wires and the contact pads of the sensor layout followed by a protective encapsulation of the solder joints, cf. Fig. 18. The encapsulation mechanically protects solder joints against breaking during sample handling and encloses the solder points to avoid direct contact with lubricant oil. The profile of the bearing is recessed by 1 mm at the contact areas. The encapsulation has a height of 0.8 mm so that it does not extend beyond the bearing running surface and impair the functionality of the bearing.

Since NiCr forms effective passivation layers on its surface, a direct soldering with conventional solder material is difficult. In order to avoid the use of strong soldering flux, an electroplating process has been developed which selectively deposits a Gold (Au) layer to the contact pads. The procedure of generating soldering contacts is depicted in the image series in Fig. 19. A polyimide mask is placed at the contact zone. The Au electrolyte is then locally applied over the mask and the deposition process is initiated by a current flow through an anode (in contact with the electrolyte) and the sensor structure which forms the cathode. After the electroplating deposition process the gold pads can be wetted with solder and wires can be attached to the pads.

For the encapsulation, an automotive grade single component epoxy compound with high temperature stability and high chemical resistance was used. The chemical stability of the encapsulation material was further tested by submerging material samples in lubricant oil and exposing them to a temperature of 60 °C for one week. For the encapsulation, a three-dimensional-printed negative mold form was attached to the bearing and filled with the epoxy compound. The sample was then degassed in a vacuum chamber to remove dissolved gas and air bubbles. Finally, the encapsulation compound was cured in a heating oven. Figure 20 shows a completed wire encapsulation. The polyimide foil was only temporarily placed as a protection of the running layer and removed for subsequent process steps. Note that a single air bubble remained in the mold during curing. However, further optimization of the degassing step eventually resulted in mostly cavity-free encapsulations.

To validate the functionality of the instrumented bearing shells, basic tests at lab scale were carried out for both temperature and strain measurements.

To test the temperature sensor, an instrumented bearing was connected to a corresponding measurement and data acquisition system and placed in a heat chamber. A PT-100 reference sensor was attached to the bearing and the chamber temperature was slowly ramped up for approximately 1.25 h from ambient temperature to 120 °C to ensure a uniform temperature distribution, cf. Fig. 21.

After this period, the heater was de-activated and the chamber cooled off for approximately 4 h. During the entire process, the resistance of one thin film temperature sensor on the bearing was recorded to establish a correlation between the reference temperature and the sensor resistance response. As illustrated in Fig. 22, a nearly linear correlation was found which features a temperature coefficient of resistance of 0.116 Ohms per Kelvin. Although the resistance change per temperature unit is comparatively small, it can be expected that with its nearly linear behavior, accurate temperature measurements can be carried out with this type of sensor.

To test the strain sensors, an instrumented bearing was clamped in a tensile testing machine (Fig. 23) to enable application of a defined force to the ends of the bearing in radial direction to compress the shell. Results of the strain sensor responses during an initial force ramp-up from 0 to 100 N are illustrated in Fig. 24. It can be found that after a short settling phase, likely caused by proper establishing of all mechanical contacts in the test setup, the bridge balance exhibits an almost linear correlation to the applied force for all three investigated sensor structures. As expected, the strain sensors at the +/−30 deg positions show a smaller bridge response compared to the 0 deg position, since the deformation due to the applied mechanical load is smaller in these areas, but the ratio of the bridge balances between the 0 deg and 30 deg positions is larger than expected based on analytical considerations (cos 0 deg versus cos 30 deg) and corresponding FE simulations. While the behavior during this first ramp-up in force looks promising, additional load cycles with linear ramps of force between 0 and 100 N come with a significant hysteresis in the sensor responses (Fig. 25). Additional detailed investigations are required and underway to investigate and understand these phenomena which currently prevent a reliable absolute strain measurement with help of the thin film sensors.

Based on the current state of work, the goal is to also test instrumented bearings on a bearing test rig and on a heavy-duty ICE to further validate the investigated thin film sensor concepts. Before such tests, however, the reliable functionality of the sensors must be proven by means of simple tests as outlined in this section, i.e., open issues such as the strain sensor hysteresis behavior have to be resolved first. Furthermore, the application of bearings in tests with rotating shafts and corresponding lubrication will likely lead to additional challenges (beyond the scope of this paper) that need to be met, such as sensor calibration and the faultless manufacturing of a second insulation layer and a running layer—parts of the bearing manufacturing process which were usually omitted for the investigations in this study.

Condition monitoring and PdM approaches for ICEs are key tools for increasing engine durability and saving resources by utilizing more of the useful lifetime of engines and components thereof while avoiding critical operating conditions caused by wear and failure. For sliding bearings, these tools require appropriate measurement technology capable of acquiring informative parameters that reflect the current condition of the bearings while withstanding the high temperatures and mechanical loads encountered inside the engine and accommodating spatial constraints. Thin film sensors were identified as an advanced instrumentation technology that can be highly integrated into the bearing (and its production process) and allows to obtain information quasi-directly at the bearing running layer while the mechanical and hydrodynamic properties of the bearing remain unaffected.

This publication presented the latest developments for an advanced thin film sensor concept for sliding bearings in ICEs that allows to monitor strain and temperature in relevant areas of interest. These parameters were chosen because they are considered both informative for bearing CM and PdM purposes as well as achievable with help of NiCr thin film sensor layers. Promising sensor layouts and positions for use in ICEs were elaborated in detail. In particular with regard to the process of manufacturing the sensor as well as the wire contacting, developments and challenges were outlined in depth. Thereby, specific issues such as sensor layer insulation, sensor resistance trimming and sensor resistance anisotropy were discussed. Eventually, basic tests at lab scale were carried out for both temperature and strain measurements to validate the functionality of the instrumented bearing shells. With the temperature sensor, a nearly linear correlation between a reference temperature and the sensor resistance response was found, which can serve as the basis for accurate temperature measurement. With the strain sensor, the bridge balance exhibits an almost linear correlation to the applied force for all three investigated sensor structures in an initial load cycle. Subsequent load cycles with linear ramps of force come with a significant hysteresis in the sensor responses, which require additional detailed investigations to investigate and understand this phenomenon which currently prevents a reliable absolute strain measurement.

The goal of further research work building on this study is to employ thin film sensor-instrumented sliding bearings on a bearing test rig and on a heavy-duty ICE to further validate the investigated sensor concepts. To achieve this, manufacturing-related challenges such as insufficient insulation of the sensor grids must be permanently overcome to enable reliable production processes of fully functional instrumented bearings. In addition, issues with the sensor functionality such as the hysteresis behavior of the strain sensor in the basic lab test must be resolved.

The authors would like to acknowledge the financial support of the “COMET—Competence Centers for Excellent Technologies” Program of the Austrian Federal Ministry for Climate Action, Environment, Energy, Mobility, Innovation and Technology (BMK) and the Austrian Federal Ministry of Labor and Economy (BMAW) and the Provinces of Salzburg, Styria and Tyrol for the COMET Centre (K1) LEC GETS. The COMET Program is managed by the Austrian Research Promotion Agency (FFG).

• Die Österreichische Forschungsförderungsgesellschaft (Award No. 865843; Funder ID: 10.13039/501100004955).

The authors attest that all data for this study are included in the paper.

Al₂O₃ =

aluminum oxide

Au =

gold (Aurum)

CM =

condition monitoring

EHD =

elasto-hydrodynamic theory

FE =

finite element

GHG =

greenhouse gas

I =

current

ICE =

internal combustion engine

LEC =

large engines competence center

NiCr =

nickel-chromium

PdM =

predictive maintenance

R =

resistance

V =

voltage