Identification of a Material–Lubricant Pairing and Operating Conditions That Lead to the Failure of Porous Journal Bearing Systems

The PJBs were first cleaned using toluene and then impregnated with the test lubricant. A standardized routine was developed to ensure cleanness of pores, avoid solvent impurities and also impregnate the bearings in a repeatable way [18]. The density of the PJBs was measured before and after the impregnation by weighing and dividing by the known volume in order to ensure that all the samples were impregnated with lubricant to at least 95%. A hole was drilled in the transverse section of the bearing for temperature measurements (thermocouple) during the tests and also for simplifying and standardized sample positioning, see Fig. 1. The hole position was intended to allow a measurement of the temperature as close as possible to the contact zone without compromising the structural integrity of the bearing. Five samples of the same material, impregnated with the same lubricant were then mounted in the tribometer benches for parallel tests. The measurement signals for evaluation of the Coefficient of Friction (CoF) were continuously recorded during the test with a sample rate of 2 kHz and statistically processed to 5 Hz time series data in the output file (one average value every 400 points), whereas wear was evaluated after the test.

PJB production is a powder metallurgical manufacturing process that consists of four major process steps: mixing, pressing, sintering, and calibrating. Mixing the metal powders together with other alloying elements ensures a homogeneous distribution of the individual materials. Thus, desired material properties can be achieved, and certain process parameters can be specifically influenced. The mixed metal powder is filled into the mold, and subsequently compacted by the press plunger. After its release from the press, the part almost has the final shape, but not yet the required strength. During sintering below the melting point of the alloying elements the part does not melt, but the powder particles merge due to diffusion processes that influence the mechanical properties of the component. Finally, calibration and embossing increases both dimensional accuracy and density while also improving the surface quality of the bearing surface.

Our PJBs were supplied by the manufacturer with an increased diameter gap (0.35%) between shaft (8 mm in diameter) and the internal diameter of the bearing, see Fig. 1a. This gap was designed to reshape the contact geometry and promote more asperity contact and consequently higher wear (accelerated tests). The spherical bearings had an external diameter of 15.6 mm and a length of 11 mm. Martensitic steel shafts were substituted after each test. The shafts had been tempered at 180 \(^{\circ }\)C for 1 h, resulting in a surface hardness of 650 ± 10 HV and of 400 HV at 0.7 ±  0.3  mm depth from the surface.

Bronze and iron-based bearing materials were used. Various quantities of graphite, copper, and molybdenum disulfide (MoS\(_2\)) were added on some configurations to improve solid lubrication. Furthermore, the effect of porosity was studied for every bearing material, with all materials being available in two different effective densities, see Table 1. The total volumetric porosity of bearings could be estimated based on the material density (inverse relation). For all materials, “low density” corresponds to 25% of total volumetric porosity, whereas “high density” corresponds to 20% of total volumetric porosity.

The tribological response of porous materials not only depends on the total porosity, but also on pore characteristics such as average pore dimensions, distribution, connectivity, and shape [20‐22]. For this work, the pore area fraction (fraction of area occupied by the pores expressed as a percentage) as a function of pore size was evaluated by optical analysis. The bearing surface images of iron and bronze-based materials at low and high density were carefully analyzed using the open-source software ImageJ [23], see Fig. 2. Three different regions (\(\simeq 3\times\) 1 mm\(^2\)) of each sample were randomly selected and analyzed to obtain statistically reliable data.

The pores of the low-density iron-based bearing (Fe-low) were generally smaller than 65 \(\upmu\)m, and the maximum fractional area (\(>2\)%) was covered by pores of 40 \(\upmu\)m size. For this configuration, there were also some pores of large dimensions (around 90 \(\upmu\)m), representing approximately 2% of the surface. Iron bearings with high density (Fe-high) exhibited lower porosity, and pores were generally smaller than those of Fe-low, as expected. Besides that, Fe-high did not present any pores larger than 60 \(\upmu\)m). Bronze bearings had a surprisingly similar total porosity and pore size distribution for both high and low density. Small pore dimensions (\(<5\) \(\upmu\)m) represent the maximum of the pore size distribution, but quite large pore dimensions (up to 105 \(\upmu\)m) were encountered in both bronze materials. The areal porosity distributions were probably so similar for Br-low and Br-high due to the calibration operation at the end of the fabrication process. The mechanical surface compression during calibration likely caused surface deformation and consequently an alteration of the surface porosity, where the bronze materials were more affected than the iron-based ones because of the lower hardness of bronze.

A wide range of synthetic lubricants was used for experimental evaluation, ranging from viscosity index of 143 (PAO) to 326 (PFPE), see Table  2.

To justify the particular bearing–lubricant combinations discussed later on in the text, we give an overview of the wear results obtained from a first high-throughput screening run prior to this study (only one sample per combination) in Fig. 1b. The headers for the rows and columns as well as the exact diametric wear values cannot be given here for confidentiality reasons, but it is clear that the combination Fe-low/PAO performs surprisingly poor when operating in the boundary lubrication regime, which is especially noteworthy since this is a standard bearing–lubricant combination. This result, which was originally believed to be either an outlier or a measurement error, led to the in-depth analysis of combinations around the region marked by the white cross.

The tribological tests were performed using our custom-built multiple bearing wear tribometer [18]. The rig consists of 5 identical test benches for performing multiple parallel evaluations, see Fig. 3. The bearings are mounted on a shaft driven by a servo motor. A frequency converter permits continuous speed variation for each servo motor. A dead weight mounted on a rod is used to axially load the bearings (pendulum), see Fig. 3. Limp spring steel strips were used for connecting the rod and the sample holder as well as avoiding axial sliding between bearing and shaft, see Fig 3b. The CoF can be measured either by evaluating the signal of the frequency converter for long-run measurements (low time resolution) or by a laser system that allows recording data at high resolution. The laser system is more accurate and used to measure pendulum deflection as a measure for the friction torque. These two different techniques for measuring the CoF were used in different stages of the tribological test, as will be explained in the following paragraphs. The reader is referred to previous publications [18, 19] for more in-depth details on the rig construction and its technical features.

Five wear tests were performed in parallel for each bearing material–lubricant combination. A constant load of 25 kg was applied in all tests to obtain a contact pressure of approximately 3 MPa. In an effort to minimize the time required for producing a measurable and widely comparable amount of diametric wear in the porous journal bearings, we tested and evaluated several speed profiles on a Fe-low bearing impregnated with a low-viscosity PAO lubricant. We compared the effect on total diametric wear of profiles with constant speed, unidirectional, as well as bidirectional saw tooth profiles featuring speed maxima of 20, 50, 150, 300, and 750 rpm. We also checked for the influence of total running time and the addition of a running-in period of 6 h at 1000 rpm. We found that including a running-in period was counterproductive, as it allowed for gentle flattening of initial asperities in the hydrodynamic lubrication regime, but such running-in does not occur in real operation. Furthermore, we found that no appreciable changes in the friction behavior could be observed after approximately 40 h of testing, and that the highest amount of wear could be produced with the lowest manageable speed. Thus, we arrived at a protocol where the speed varied according to a bidirectional saw tooth profile with a speed maximum of 20 rpm in each direction with a period of 10 s, see Fig. 4a. Each test ran for a net duration of 44 h in this “high wear regime”. The saw tooth speed profile was interrupted by Stribeck ramps with exponential speed increase up to 3000 rpm (see Fig. 4b) after fixed time intervals: every hour up to 5 h of total testing time, then every 3 h up to 20 h, and finally every 8 h until completion of the test, yielding a total of 13 interruptions of the saw profile (after 1, 2, 3, 4, 5, 8, 11, 14, 17, 20, 28, 36, 44 h of testing). Each Stribeck curve was recorded three times in a row to ensure statistical reliability. Selected bearing material and lubricant combinations were also evaluated using a speed profile of 50 rpm for appraising a sliding speed effect on tribological behavior. In particular, the speed dependence of wear in the combination of Fe-low with PAO was of high interest, as this pairing performed so poorly at 20 rpm.

The coefficient of friction (CoF) was continuously measured during the saw tooth speed profile phases via the signal of the frequency converter, where only the values at ±20 rpm were compared, leading to a sampling frequency of 5 Hz. During the Stribeck ramps, the CoF was recorded at a sampling frequency of 2 kHz via the laser instrumentation that measures the pendulum deflection. Wear was measured using a tactile two-point comparison instrument (Schwenk OSIMESS) after the end of the test in the central and the near-border regions (approximately 1 mm from external border) of the internal bearing diameter in the inner bearing bores [24].

Figure 5 shows a typical time development of the CoF over the 44 h testing time, with the Stribeck ramp data removed and the respective interruptions clearly visible as CoF jumps. The CoF spikes in the graph represent test interruptions in the saw speed profile for performing the exponential speed profile of the Stribeck curves. These speed interruptions caused start-and-stop issues in the contact (starvation), and therefore direct asperity contact and consequently a local CoF increase (spikes). Furthermore, the high speed during the Stribeck profile increased the temperature (see Fig. 10), which led to a viscosity drop and lubricant thinning, subsequently promoting asperity contact and a transient CoF increase. Following the temperature spikes caused by high speed during the Stribeck curves, the temperature slowly decreased to its previous (ambient) level, see (see Fig. 10a–d). The temperature drop then caused the lubricant viscosity and thus the load support capacity to increase, consequently reducing asperity contact. For these reasons, the CoF decreased in a similar way as the trend of the temperature, see Figs. 5 and 10b). For all the material and lubricant combinations, we observed that the CoF reached a steady value after a run-in period. This behavior is mainly due to changes in the contact geometry during the tribological test. Several cycles between the contacting bodies (shaft and inner bore of the bearing) caused asperity flattening, and thus the contact pressure was reduced and a lower CoF was obtained. The frictional response over time of the high-density steel bearings (Fe-high) tested with reference oil (PAO) is presented as an example for such behavior in Fig. 5. In this case, the CoF assumed a steady value after roughly 30 h of testing.

The studied material and lubricant combinations showed different levels of CoF, but the general trend was the same (CoF spikes due to test interruptions, running-in in the beginning with higher CoF, and steady CoF state at the end). For these reasons, a general comparison between all sample combinations was made based on the steady CoF state at the end of the test (44 h). Coefficients of friction for each bearing material and lubricant combination were calculated as average of the five identical tests run in parallel on different benches. The main results for the CoF of the steady state are visualized in Fig. 6. Due to the wide range of studied conditions, the results were grouped according to material, lubricant, and speed variation for improving the interpretation and summarizing the main outcomes, see Fig. 6.

A bearing material variation (Fig. 6a) was evaluated using the synthetic PAO as a benchmark lubricant and a saw profile speed of 20 rpm for all bearing materials. For the bronze bearings (Br, brown color series in Fig. 6), we found that the addition of graphite (BrGr) slightly improved tribological performance, whereas significant CoF reduction was achieved by adding molybdenum disulfide (MoS\(_2\)–BrMo), as expected [25‐27]. The two BrMo systems consistently exhibited remarkable running-in behavior, with a drop in the CoF at 20 rpm from 0.35 to below 0.1 within 8 h of operation in the case of BrMo-low. As will be mentioned in the next section, these particular systems were also prone to noticeable, but contained changes in the bearing diameter, so the good tribological performance may be a result of a redistribution of MoS\(_2\) that was loosely bound in the bearing matrix to the sliding interface, which is accompanied by the smearing shut of pores in the contact region. On the other hand, the addition of a substantial amount of copper (20%—FeCuGr) and graphite (FeGr+) to the iron-based bearing (Fe) brought drawbacks. The increase of copper together with the addition of graphite (FeCuGr and FeGr) increased CoF compared to the reference iron-based material (Fe). However, the worst performance was observed for the material with reduced copper and the maximum achievable addition of graphite (FeGr+), probably because not only the chemical composition was changed but also the density. In this case, FeGr+ bearings exhibited generally lower density than reference iron material Fe, and therefore this behavior could be caused by higher porosity or by lower hardness. It was shown in [20‐22, 28] that the increase of porosity and pore dimensions generally leads to an increase in friction. Large pores can cause lubricant film breakdown with the consequent increase of asperity contact and CoF [21, 22]. The graphite addition in FeGr+ was significantly higher than in the other compositions, especially considering the volumetric content. Since here the graphite is only loosely bound within the bearing matrix, a significantly higher volume of graphite in FeGr+ caused more unsteady tribological behavior.

Comparing the same bearing material with different densities (low- and high-density series in Fig. 6a), bronze-based bearings showed only small differences, namely that in general low-density materials (Br-low) exhibited lower CoF values than those with high density (Br-high). However, compared to the differences between the respective results, the error bars on the bronze CoFs are too large for claiming that these differences are statistically significant, and the areal bearing surface porosities between Br-low and Br-high are almost indistinguishable, see Fig. 2e–h. We encountered the opposite behavior for iron-based materials, where low-density materials had increased friction compared to high-density ones. This behavior is again in agreement with [20‐22, 28], where it was demonstrated that an increase in porosity and pore size (see the iron-based series in Figs. 6a and 2a–d) also increases the CoF. It should also be noted that Fe-low produced highly scattered friction data in combination with PAO, the reasons for which will be discussed in the next sections.

The lubricant variation (Fig. 6b) was evaluated using the high-density bronze (Br-high) and iron-based material (Fe-high), using a saw profile speed of 20 rpm. These two bearing materials were chosen in this comparison for their high industrial relevance, as they are frequently used standard components that should be compatible with a wide range of lubricants. The bronze bearings (high density) generally produced higher friction values than iron ones for all tested lubricants, see Fig. 6b. Looking at the results for the bronze-based bearings, it was also observed that the lubricant variation did not lead to significant CoF changes. For iron-based materials, however, the best results were obtained with the PAO lubricant, which features low viscosity and a low viscosity index.

Finally, a speed increase of the saw profile (from 20 to 50 rpm) was evaluated (Fig. 6c) only for low-density iron-based material (Fe-low) with ester and PAO lubricant. It was observed that when using ester, the speed variation did not significantly affect the steady CoF of Fe-low. However, higher speeds (50 rpm) significantly reduced the CoF when using the PAO lubricant. Here, we again mention the high CoF variation for Fe-low tested at 20 rpm with the PAO lubricant (see also material variation, in Fig. 6a). Reasons for that will be discussed in the next sections.

The total wear for each material was expressed as average values of the measurements taken at the middle and near the two edges (approximately 1 mm from external border) of the inner bore diameter of the bearings. Furthermore, the value obtained was averaged between the five identical samples tested on parallel benches. The main results, grouped according to speed, bearing material, and lubricant variation, are shown in Fig. 7. Before going through the results, it should be mentioned that a diametric wear below 10 μm can be considered contained, while values below 5 μm are considered an expected effect of running-in and can therefore be neglected.

Generally, the bearing materials tested with different speed and lubrication combinations did not wear out significantly after the 44 h tests, with the exception of some cases (Fe-low and BrMo-low, PAO lubricant at 20 rpm), see Fig. 7a. The low-density series of bronze with MoS\(_2\) (BrMo-low) exhibited contained wear (below 10 μm), whereas the iron-based bearing (Fe-low) produced unacceptable wear (more than 60 μm averaged over 10 tested bearings).

For BrMo-low, the “higher” wear (between 5 and 10 μm) could be due to higher porosity (lower density material) and due to the solid lubricant MoS\(_2\). It was proved in previous works [29‐31] that an excessive quantity of solid lubricant may deteriorate mechanical material properties and decrease wear resistance. Furthermore, it should be considered that solid lubricants are primarily added for extreme starved lubrication conditions and they may not work well in combination with some contaminants, e.g., hydrocarbons, water, and oxygen [26, 32]. In the case of Fe-low, the bearing clearly failed during tests, the possible reasons for which will be evaluated in more depth in the next section.

When comparing various lubricants in high-density Fe and Br bearings, as shown in Fig. 7b, it can be stated that wear is practically negligible here. The slight trend that Fe wears more than Br is only broken for PFPE-lin, but the absolute values are so low and the data scatter for the Br system is relatively large that a serious discussion about the causes for this behavior is not possible.

It is noteworthy that the increase of speed (from 20 to 50 rpm) for Fe-low tested with PAO reduced the diametric wear to acceptable values, see Fig. 7c. The increase of speed could bring the system to a less severe regime of lubrication, therefore avoiding direct asperity contact and excessive wear.

This section will focus on Fe-low tested with PAO at 20 rpm, since this combination exhibited behavior distinctly different from all other combinations both in terms of CoF and wear response, see Figs. 6 and 7. Most specimens exhibited “regular” friction curves, as shown in Fig. 5, with an initial run-in, CoF spikes at expected time intervals due to test interruptions for Stribeck ramps, and finally a steady CoF state. For Fe-low (with PAO at 20 rpm), the CoF results did not follow this trend, see Fig. 8a and b, where the friction curves of the 5 parallel benches were plotted for Fe-low and compared to the “regular” curves obtained for Fe-high under the same testing conditions (PAO lubricant using 20 rpm saw speed profile). The results of the 5 parallel benches were then averaged to scrutinize the overall friction trend, see Fig. 8c. Apart from producing significantly higher CoF than Fe-high, the Fe-low samples exhibited a greatly increased standard deviation in the CoF results, see Fig. 6a.

The friction coefficient was also evaluated during the Stribeck ramps after fixed testing times (13 in total, see Sect. 2.3). Stribeck curves were recorded three times in a row during every test interruption, where the data of the first one were discarded and the averaged values of the last two can be found in Fig. 9a and b. Note that Fe-low samples showed a typical Stribeck curve shape [33] at the very beginning of the test (after 1 and 2 h), whereas after that the friction curves assumed unstable values during the exponential speed increase, see in Fig. 9a. This behavior is probably due to lack of lubricant film between the shaft and the inner bearing bore, in fact the CoF at high rotational speeds assumed similar values to those at the lowest speeds in the boundary lubrication regime. Higher values of friction at high speed (\(\ge\) 1000 rpm) indicate more severe contact conditions and consequently higher wear rate, as observed in Fig. 7. On the other hand, Fe-high produced typical Stribeck curves in all the tests, see Fig. 9b. The Stribeck curves for Fe-high assumed lower and lower values with longer test duration, from a run-in stage a more steady one, see Fig. 9b, with the minimum CoF moving towards smaller speeds with increasing test duration. By contrast, for the Fe-low samples, the friction values in Stribeck curves generally increased over time and the minimum shifted towards higher speeds, indicating lubrication failure, see Fig. 9a.

The friction results of the Stribeck curves were also visualized as 2D heat maps. The vertical axis corresponds to the overall test time from top to bottom, and the horizontal one to the rotational speed, see Fig. 9c and d. The speed axis was plotted using a logarithmic scale to highlight the differences in terms of CoF especially at low rotational speeds. Note that these maps only represent friction values during the Stribeck ramps and not the overall frictional behavior at 20 rpm. The heat map was drawn as interpolation of friction curves and visualizes their development over time. Results from Fe-high and Fe-low were plotted using the same color scale to facilitate comparisons. From the heat maps, it becomes clear that Fe-low assumed low friction values at the beginning of the test, whereas friction subsequently increased during the test (vertical axis) and then also for higher rotational speeds (horizontal axis). On the other hand, Fe-high exhibited lower friction values during test and speed variation.

The temperature in the bearing was continuously measured at a sampling rate of 5 Hz during tests using a type K thermocouple. Results for iron-based materials (Fe-low and Fe-high) using PAO lubricants (PAO and PAO+Gn) are shown in Fig. 10a–d. It is possible to see a similar trend, namely that the temperature assumed steady values during the saw speed profile, as well as temperature spikes caused by the higher rotational speeds during the test interruptions for the Stribeck curves. Note that all four graphs in Fig. 10a–d were plotted using the same scale in the vertical axis (temperature); therefore, it is evident that temperature spikes reached higher values for Fe-low than for Fe-high with the same lubricant, probably due to lack of lubrication as explained in the previous section.

The wear measured at the end of the test was plotted over the maximum temperature reached during each experimental test, see Fig. 10e–g. From the graph, all experimental results could be divided into two groups depending on the relation between wear behavior and maximum temperature reached. The majority of the experiments lay on a wear plateau, where the increase of temperature (up to roughly 60 \(^{\circ }\)C) did not significantly affect the wear behavior. By contrast, several experiments exhibited a linear correlation between maximum temperature and wear, where temperatures exceeding 60 \({}^{\circ }\)C led to unacceptable wear. It turns out that only the iron-based materials with low density (Fe-low), tested with PAO lubricants (PAO and PAO+Gn) at 20 rpm, belong to the latter group. When plotting the maximum temperature over the coefficient of friction for the two groups (Fig. 10f), although the data are quite scattered, it becomes clear that group 1 (blue) has a weak dependence of the temperature on the CoF, which is indicative of the corresponding systems achieving hydrodynamic lubrication. Here, the heating of the system is dominated by energy dissipation within the lubricant due to its viscosity. The slope of the CoF-vs-T relationship for the systems in group 2 (red) is almost 6 times as high as for group 1, which is evidence for operation in the mixed lubrication regime. Here, intermittent direct contact between asperities of the shaft and the bearing leads to plastic deformation, releasing higher amounts of energy, heating the system, thus thinning the lubricant and further reducing its load bearing capacity. When these deformation events become too severe, they result in wear and material loss. It may therefore be fallacious to assume that elevated temperature under mixed lubrication conditions causes wear, or that wear in the mixed lubrication regime causes the temperature to rise. Rather, both are correlated results of the direct asperity contact, which in turn are likely caused by a compound effect of a limited load bearing capacity of the lubricant at the low sliding speed combined with the high porosity of the bearing. Another possible explanation for the observed correlation may lie in the lubricant shear stress having a detrimental effect on the more fragile porous structure of the low-density bearings. This would lead to elevated amounts of debris accumulating in the lubricant, which may increase the friction and thus lead to the temperature increase.

Finally, plotting the diametric bearing wear over the CoF at the end of the test allows us to elucidate the common misconception that a high coefficient of friction automatically also leads to high wear. What Fig. 10g shows rather clearly is that in the overwhelming majority of cases, bearing wear is constantly below the negligibility threshold independent of the CoFs ranging from 0.04 to 0.24, and that only for the high-temperature cases in the mixed lubrication regime, a distinct correlation between the two quantities is apparent.

To sum up, all the material and lubricant combinations led to acceptable tribological results, apart from Fe-low tested at 20 rpm with PAO oils. A possible hypothesis could be based on the tribological effect of surface porosity on this particular combination. Even if Fe-low (high porosity) features lubricant reservoirs and good lubrication supply, it seems as if the presence of surface porosity combined with PAO lubricants brought drawbacks in terms of pressure build-up, thus resulting in less load carrying capacity, especially at elevated speeds.