Optimised PVDF Placement Inside an Operating Hydrodynamic Thrust Bearing

Abstract

In our previous work we demonstrated the feasibility of using Polyvinylidene Fluoride (PVDF) sensors inside an operational thrust bearing and were able to measure the blade passing frequencies (BPF) due to an asymmetric flow around different propellers. In that work however the sensors were positioned inside the flat surface of the stationary portion of the bearing with the tilted pads rotated on the opposite side. Due to this configuration the output signal of the PVDF consisted of a superposition of the pad passing frequency (PPF) and the blade passing frequency (BPF) making it difficult to extract useful information from the results. Here, an improved bearing pad-film configuration is proposed in order to minimise the effects of the PPF. By embedding the films inside the pads, positioned on the stationary side of bearing, and rotating the flat surface, it was possible to eliminate the PPF and significantly increase the signal to noise ratio. The measured results give a better understanding of the fundamental vibratory components that arise from the propeller-shaft system.

1. Introduction

Most marine vessels have propellers at the stern operating in non-linear wakes. This propeller placement is done primarily for ease of construction and propulsion efficiency. The main disadvantage of such a setup is having to operate in a region where the wake is non-linear. This generates noise from an oscillating thrust and, in terms of hull induced vibrations at frequencies below 100 Hz [1], can have large effects on passenger comfort, mechanical wear and acoustic detection. Developing an accurate vibrational model of the propulsion system in a non-uniform wake is extremely difficult. It is necessary to have a good understanding of the hydrodynamic forces inside the thrust bearing and measuring this force is logistically difficult. This is due to the complex nature of thrust bearings which self stabilise based on operating parameters such as speed and load [2]. To simplify the inclusion of thrust bearing stiffness in analytical modelling, simple linear stiffness representations are used [1, 3, 4], however these are limited as it has been shown that due to large perturbations of the oil film thickness, a nonlinear response is required [5, 6, 7]. Therefore measuring the internal hydrodynamic forces inside the thrust bearing is ideal. This paper attempts to address this problem by using intrinsic piezoelectric sensors installed inside the thrust bearing itself. Conventionally this would be done with accelerometers however their physical properties make them unsuitable for use inside a moving bearing. Another option is the use of thin piezoelectric polymer film such as PVDF (Polyvinylidene Fluoride).

In our previous paper [8], an in-depth review of the use of PVDF to measure dynamic forces in static structures [4, 9, 10, 11, 12, 13, 14, 15, 16] was discussed. It was noted that PVDF cannot measure absolute pressures, only the fluctuations, which is a common use of PVDF in fluids [17, 18, 19, 20, 21, 22], due to its piezoelectric properties [23, 24]. The paper reported on the use of PVDF inside a thrust bearing to measure the pressure fluctuations and infer the change in contact force. Due to limitations in the experimental design, namely the placement of the PVDF films in relation to the pads on the rotating body, the signatures captured from the PVDF films were heavily influenced by the Pad Passing Frequency (PPF). Hence, although that study found that the sensors could measure the pressure fluctuations generated by the sliding pads, the true nature of the excitation of interest, the BPF was masked.

Here, the design of the experiment was reworked so that the true excitation force from the propeller and the resulting change in contact force within the bearing could be observed without the influence of the PPF. This was achieved by placing the PVDF in the tilted portion on the stationary body.

The previous configuration (Figure 1a), resulted in the pressure profile generated by the pad sliding over the PVDF film. This led to the dynamic pressure fluctuations being subjected to the PVDF from the pad motion relative to the PVDF sensor. The new configuration proposed in this study (b) removes this component and results in the PVDF film being subjected to the dynamic fluctuations arising from the shaft axial load and hence propeller. The static component of the pad’s pressure profile cannot be observed in the signature as a result of the intrinsic nature of PVDF.

Figure 1.

The arrangement of the PVDF film within the thrust bearing in our previous work [8] (a) and in this work (b). In both arrangements the PVDF is embedded on the stationary body and ‘v’ indicates rotational direction of the rotating body.

2. Materials and methods

The original rig [8], shown in Figure 2, is used here and its main attributes described here for completeness. It was designed to model a typical propulsion system used in marine settings. The system consists of a propeller (prop), shaft, journal bearings, and a thrust bearing attached to a supporting plate structure which represents the hull. The propeller provides axial thrust to the shaft which is transmitted to the hull (supporting plate) via the thrust bearing. The coupling, known as the “dog clutch”, is rigid in torsion but flexible in bending thus allowing axial motion to be transmitted along the “floating” shaft. This helps to isolate any unwanted vibration from the motor. The 10 mm thick, 1500 mm wide supporting steel plate is fixed to a concrete block to take the thrust load generated by the propeller and also supports the motor as shown.

Figure 2.

Schematics of the experimental set-up: (a) layout of entire physical system; (b) enlarged view of propulsion system.

A two-blade and three-blade propeller of equal diameter (0.22 m), shown in Figure 3, were separately coupled to the propulsion system in the water tank (0.99×0.59×0.58 m) to observe different blade passing forces. A larger (diam = 0.3 m) alternative style bronze two-blade propeller was also used (Figure 3).

Figure 3.

Propellers used in the experiment: (a) standard two-blade; (b) bronze two-blade; and, (c) three-blade.

In addition to the four test configurations (No prop = NP, standard two-blade prop = TBP, bronze two-blade prop = TBBP and three-blade prop = ThBP), two PVDF arrangements were investigated with these test configurations. For each case, the non-rotating component into which the PVDF was embedded was constructed out of ABS (Acrylonitrile Butadiene Styrene) using a 3D printer. Since the forces, temperatures and operating times were relatively low, it was presumed that using the plastic material inside a functioning, small-scale, thrust bearing would not be a problem. This was confirmed by our previous study [8].

Two different designs were investigated regarding the placement of the PVDF: Case 1 - the stationary component was a plain washer with the thrust washer in the rotating component sliding over top as in our first study [8]; Case 2 - was arranged such that the stationary component had the thrust washer and the PVDF embedded on top, with a plain washer in the rotating component sliding over top. Both arrangements (cases) had the PVDF fixed to the non-rotating body to permit the signals to be measured without the need for telemetry or slip rings (Figure 4).

Figure 4.

Thrust bearing comprising stationary and rotation components (a). Stationary bearing configuration: Case 1 - PVDF embedded in plain washer (b) and Case 2 - PVDF embedded in thrust washer (c).

In Case 1, (Figure 4), where the stationary component was the plain washer, recesses were made in the washer to insert two PVDF films as shown. Opposite the plain washer, a thrust washer, also constructed out of ABS, was used with two pads angled at two degrees to the surface. In Case 2, in Figure 4, the thrust washer was printed with two, two-degree tilted pads and the PVDF film was embedded into the tilted portion of the washer. Opposite the thrust washer, a plain washer was used in the rotating component. This configuration better resembles what is seen in a typical thrust bearing in a marine propulsion system.

The PVDF sensors were cut from a sheet of 110 μm thick film obtained from Measurement Specialties. Contacts were made using conductive copper tape on the electrodes and the films were glued using West Systems 105 epoxy resin inside the two recesses in the plastic washer. The PVDF sensors were protected by covering the interacting surface of the washer with a thin layer of epoxy. Heat from a blow torch was briefly applied to the epoxy overlaying the PVDF sensors to remove any small bubbles in the thin layer before it was left to cure. Connectors were then added to the end of the cables from the PVDF sensors such that they could be operated independently.

A Brüel & Kjaer (B&K) 2635 charge amplifier was used to condition the signal from the PVDF film. In conjunction with the B&K Pulse Time Data Recorder software, a B&K LAN-XI data acquisition system (Type 3050-B-060) was used to analyse and collect data in the time domain from the PVDF. At a sampling rate of 16,384 Hz, any higher frequency noise from the sensors would be collected while retaining manageable file sizes.

The data was post-processed using Matlab. The raw voltage measured from the PVDF film were converted to force via the measured calibration factor as discussed in our previous papers [8, 25]. Although no physical trigger was implemented in the experimental setup, the time domain signals were aligned using Matlab’s “alignsignals” function. The frequency-domain data was obtained using the inbuilt FFT function by dividing each track into 8 equal segments to obtain a better frequency representation of each time-domain signal. Each segment was windowed with a Hamming window, with 50% overlap to reduce the effect of windowing.

The set-up enabled the prop shaft to be spun in different configurations. The first being with no propeller at the end of the shaft to give a representation of the noise transmitted to the sensors from just the shaft spinning. Then, three different propellers were attached to the shaft in turn and rotated at speeds of 0 to 600 RPM in 20 RPM increments. Each configuration was performed for both Case 1’s and 2’s design.

3. Results and discussion

Figures 5 and 6a shows the results with Case 1 bearing arrangement at 500 RPM for each of the three aforementioned propellers. The features seen here are representative of what is discussed in our previous paper [8]. Figures 5 and 6b show the result when the bearing design of Case 1 (Figure 4) is replaced with Case 2 (Figure 4). The result pertains to the force from one of the embedded sensors at 500 RPM, the other channel shows similar trends. In this arrangement, the sensor signature is dominated by the blade passing effects.

Figure 5.

Time data of the captured dynamic force from the PVDF with Case 1’s design (a) and Case 2’s design (b). The two (TBP), bronze two (TBBP) and three (ThBP) blade propeller is spun at 500 RPM when compared to no propeller (NP).

Figure 6.

Frequency data of the captured dynamic force from the PVDF with Case 1’s design (a) and Case 2’s design (b). The two (TBP), bronze two (TBBP) and three (ThBP) blade propeller is spun at 500 RPM when compared to no propeller used (NP).

The time-domain signatures in Figure 5 is difficult to interpret due to superposition of the frequency components contained in the signal, however, it can be seen that the propeller force varies approximately ±2 N in Case 2’s result. This change in propeller force is similar in magnitude for the three propeller types. In addition to this, the plain shaft exhibits a large signature at the shaft rate, which is attributed to the shaft rotating in the journal bearing which is coupled to the PVDF film via the thrust bearing.

It can be seen in the frequency domain of Figure 6 that the most significant peak after the shaft rate $\left({\mathrm{A}}_{f1\mathit{\text{Case}}2}\right)$ is the second harmonic $\left({\mathrm{A}}_{f2\mathit{\text{Case}}2}\right)$ for the two-blade result and the third harmonic $\left({\mathrm{A}}_{f3\mathit{\text{Case}}2}\right)$ for the three-blade propeller, these correspond to the respective BPFs. The effect of the PPF is eliminated as ${\mathrm{A}}_{f2\text{Case}2}<<{\mathrm{A}}_{f3\text{Case}2}$ with reference to the three-blade result; the improvement is less apparent if considering just the form without accounting for the amplitude regarding the two-blade data as the PPF and BBF are superimposed for these arrangements involving a bearing comprising 2 pads and propellers with 2 blades. This is in contrast to when the PVDF is embedded as in Case 1’s design. No matter which propeller is used, the second harmonic of the shaft rate is the most significant, ${\mathrm{A}}_{f2\text{Case}1}>{\mathrm{A}}_{f3\text{Case}1},$ due to the two pads sliding over the PVDF film. This artefact is a by-product of the set-up and confounds the true nature of the propeller excitation forces, as revealed in Figures 5 and 6 (L2, L4).

These trends can be further demonstrated across all test speeds, and more clearly depicted in the frequency domain (Figure 7). For the plots shown on the left in Figure 7, 10 s of measured data was recorded for each rotational speed tested. The 10 s records were stitched together to produce a spectrogram. For clarity, only the three blade propeller (ThBP) results are shown (similar results were observed for the two blade propeller variants). The spectrogram amplitudes along the second and third harmonics have been extracted and are shown in Figure 7 and denoted by $A_{f2}$ and $A_{f3},$ respectively. This method of data representation clearly evidences that the second harmonic is typically lower or much closer to the third harmonic for the Case 2 arrangement compared to the Case 1 arrangement.

Figure 7.

The frequency response of the PVDF sensors across all test speeds for the Case 2’s arrangement and Case 1’s arrangement (a) with the three-blade propeller (ThBP). The magnitudes at the second and third harmonics are shown in (b) for both test cases.

With reference to the Case 1 results depicted in Figure 7a, a bifurcation is observed about a 50 Hz centre frequency. (This phenomenon was also observed for the TBP tests not shown here). The bifurcation was not observed in any of the Case 2 tests. The 50 Hz centre frequency is attributed to background Australian supply voltage EMF. The bifurcation is believed to result from the superposition of the strain in the PVDF arising from the hydrodynamic force which moves with rotational speed and the strain caused in the PVDF as a result of it being subjected to the external field. Further investigation is required to understand this phenomenon.

The proposed PVDF arrangement of Case 2 can also be compared to the initial PVDF bearing pressure measurement investigation presented by the authors [25]. In that work the PVDF was embedded in the flat base of the thrust bearing similar to Case 1’s configuration presented herein. The vertical test rig arrangement [25] facilitated different rotational speed and different fixed bearing gap investigations in the absence of a fluid immersed propeller. Figure 8 shows 300 RPM data from that study processed into the frequency domain and compares the result to that of the current study for the Cases 1 and 2 arrangements for the same speed with no propeller (NP). As the sensitivities of the sensors vary, the data has been normalised to the shaft rate. There are two attributes: (1) that the second harmonic is reduced when compared with the original study; (2) that the noise floor is reduced across all frequencies, that warrant more detailed explanation. Firstly, the results obtained by [25] and Case 1 here show an increase in the second harmonic of the shaft rate when compared to newly proposed Case 2 arrangement. The larger amplitude pertaining to the original [25] study compared to Case 1 here is attributed to the increased pressure formed under the tilted pad as a result of the original set-up which forced a fixed film thickness whereas in the current study the bearing can adjust the film thickness to achieve force equilibrium. Secondly, and of potentially greater importance, is that the noise floor across all speeds has been reduced significantly using the Case 2 arrangement. It is thought that this can be attributed again to the new arrangement’s reference frame. Specifically, that unlike the arrangements that have the pad moving passed the PVDF, because the PVDF films are mounted on the pads in Case 2 then they effectively experience the cleaner established pressure field above the pad and are not as influenced by the frequency content contained in the upstream and downstream flow and eddies [25] on each side of the pad. Whilst further experimental and numerical work is underway, these results add credence to the utility of the new arrangement.

Figure 8.

Comparison of the [25] frequency content to that of Cases 1 and 2 at 300 RPM with no propeller.

4. Summary

The focus of the paper is to evidence differences between results from PVDF sensors when they are placed within different surfaces of the thrust bearing. In particular, the pad passing frequency (PPF) was discovered when the rotating pads act across the surface of the sensors. Also observed was an increase in broadband noise associated with fluid motion generated by the rotating pads. Both of those observations are important for the measurement of the interaction force within a thrust bearing. The PVDF placement on the pad, on the other hand, provides signals free from the pad passing and broadband signals as a result of the changed reference frame. This significant reduction in the noise floor across the full frequency of rotational speeds tested is achieved without post processing filtering. Despite its promise, this latter configuration of the PVDF placement is by no means to be the optimal placement, as the interpretation of the PVDF output in terms of the force components on its surface and the dynamic response of the structure upon which the sensor is attached still require more sophisticated experimental and modelling work.

Conflict of interest

The authors declare no conflict of interest.