The results of the FEA include the deformation and loading analysis for the final setup of the iterative study and the data of the sensitivity analysis. Next, the experimental results are presented along with data from the FEA for validation.
3.1. Results of Finite Element Analysis: Deformation, Loading, and Sensitivity
For a general understanding,
Figure 5 shows the results of the FEA exemplarily for the single element configuration (final model after the iterative study). The deformation behaviour can be characterised by the displacement of point
C, representing the ball’s centre of rotation. It moved distally towards the embedding resin (
= 0.48 mm) and forced the proximal artificial bone segment to bend in the lateral and posterior directions (
Figure 5a). As for our previous work, the stress distribution showed a concentration at the cavity base, which was also the global maximum for the metallic implant component (σ
piez = 273.5 MPa). The piezoelectric element was relatively homogenously compressed, apart from the top and bottom end face, where influence of the contact boundary conditions resulted in a local stress rise (please note the scaling of the legend for
Figure 5c). This effect faded after a few elements. For the undisturbed midplane, the stress σ
piez amounted to 16.6 MPa. The electrical interconnection and polarisation of the piezoelectric element’s layers resulted in the alternating distribution of the electrical potential. The simulated open-circuit voltage V
OC was 11.6 V.
The output parameters resulting from the iterative study are shown in
Table 5. More than 150 different design points were investigated to successively minimise the differences to the output values of our previous study [
41]. The finally chosen loading situation was also used to simulate the stacked configuration. Apart from the stress in the piezoelectric element σ
piez, the aimed difference of ≤
was achieved. For the stacked configuration, the results were even better, with relative differences of ≤
. The stress in the piezoelectric element’s representative undisturbed midplane σ
piez was higher for both configurations compared to the previous work (17.7% for the single element and 7.1% for the stacked configuration) and the aimed value of ≤
was not achieved. No further decrease could be realised without impairing the results of the open-circuit voltage V
OC or the contact forces F
33. However, the absolute values for the deviation were small (<3 MPa).
The scaling of the force components for the final configuration is shown in
Table 6. To reproduce the loading situation from the previous work, the absolute values of the x- and y-component needed to be strongly reduced. In contrast, the z-component was slightly increased. However, the difference between the total force values was only 1.8%.
The maximum power output calculated for the single element amounted to 18.4 µW for a resistance of 1.35 MΩ and 72.7 µW for a resistance of 0.68 MΩ for the stacked configuration. The calculated voltage slope for the two gait cycles and the power for different total load resistance are shown together with the corresponding experimental data further below for the stacked configuration (Figures 8–10).
The results of the sensitivity analysis are described below, revealing the relevant input parameters influencing the output. The full result data are shown in
Figure A1,
Appendix A.1. Regarding the experimental validation, the strain gauge strain
, the displacement of point
C in vertical direction
and the power output were evaluated. In the sensitivity analysis, the simulated strain gauge strain
was unaffected by nearly all the parameters but the implant material and the alignment. The ±10% change in the Young’s modulus of the implant led to an equal change (–9 to 11%) of the strain
. The higher Young’s modulus resulted in a lower strain and vice versa. The influence of rotation in the antero-posterior plane was lower (–0.6 to 1.4%) compared to the change of –8 to 7% for rotation in the medio-lateral plane.
The displacement of point C in vertical direction was the only parameter slightly sensitive to the change in the embedding level (±1%). In a comparable range, the influence of the cement mantle thickness and resection level could be found. The variation in Young’s moduli of the artificial bone and implant of ±10% resulted in a change of ±4 to ±5% of . With –25 to 66% for the antero-posterior rotation and –20 to 51% for the medio-lateral rotation, the alignment had the most pronounced sensitivity towards .
The power output was calculated on the basis of the contact force F
33 acting on the piezoelectric end faces, which is therefore presented first. The results of the sensitivity analysis show that the contact force F
33 and the open-circuit voltage V
OC were identically influenced; therefore, only the contact force F
33 is presented and shown in
Figure A1,
Appendix A.1. The 10% change in the stiffness of the implant or UHWM-PE housing changed the contact force in a range of up to 10%, respectively up to 6%. Lower stiffness of the implant and higher stiffness of the housing increased the contact force F
33. The resection level more distally (i.e., the energy harvesting position closer to the cut) increased the loading (by around 4%). A more proximal resection level decreased the force level by up to –5%. With –6%, a thinner cement mantle showed a relevant sensitivity, and a higher thickness resulted in a rise of +2.5%. The contact force F
33 was smaller (–7%) for a higher angle in the antero-posterior plane and vice versa. The highest sensitivity for this parameter was for the rotation in the medio-lateral plane with ±13%. The sensitivity of the power output for each configuration was twice as high compared to the sensitivity to the force F
33. The change in the artificial bone’s material data had an influence of around ±2%, for the UHWM-PE housing it was ±12%, and for the implant up to 19%. Different resection levels changed the power output by up to –10% (more proximal) or 9% (more distal). A thinner cement mantle reduced the power by 12% and a thicker geometry increased it by 6%. The influence of alignment was again most pronounced, with –13 to 8% for rotation in the antero-posterior plane and –25 to 27% in the medio-lateral plane.
For the stress concentration in the cavity base σimp, changes in the material’s Young’s moduli by ±10% resulted only in changes of around ±1% or less. A slightly higher sensitivity was shown by changing the resection level: With a +2 mm more proximal cut, the stress σimp decreased by 3%. In contrast, a rise of 1.5% resulted from resecting more distally and nearly directly above the cavity. The change in the cement mantle geometry by 1 mm resulted in a stress rise of 2% for a thicker cement mantle and a decrease for a thinner cement mantle. With 8–12% of relative change, the rotation in the medio-lateral plane and antero-posterior plane had the highest influence on the loading in the cavity base, with a maximum rise by 32.0 MPa to a total value of 301.6 MPa.
Regarding the changed material properties, the stress in the piezoelectric element σpiez was influenced by the Young’s modulus of the implant (up to 10%) and the UHMW-PE housing (up to 7%). A decreased stiffness of the implant and an increased stiffness of the UHMW-PE raised the stress. A resection level more distally increased the stress σpiez (by 8%), whereas the stress deceased for the thinner cement mantle and the more proximal resection levels by 7%. The highest sensitivity was shown for the rotation in the medio-lateral plane (±12%), with a maximum value of 20.4 MPa. In the antero-posterior plane, it amounted to up to 6% only. In the antero-posterior plane, the effect of rotation on the stress σpiez was inverse to the change in the contact force F33. Although a higher angle increased the stress σpiez, it reduced the acting contact force F33 and vice versa.
3.2. Experimental Results and Validation
The implantation result was analysed under X-ray (X-ray unit: Gierth HF80ML Ultra Light, GIERTH X-Ray international GmbH, Riesa, Germany; digital detector system: Leonardo DR 1210, Oehm und Rehbein GmbH, Rostock, Germany) and compared to the CAD model. The cross-sectional view showed good accordance (see
Figure 6). In particular, the position of the energy harvesting system to the resection level matched well under visual inspection. The overall cement mantle geometry was consistent with the CAD model; however, in the proximal total hip stem region, it was slightly thinner for the experimentally realised implantation. A portion of the lateral implant surface was uncovered by the cement mantle and was in direct contact with the cortical geometry for both the virtual and the realised implantation.
The averaged measured maximum strain
at the maximum force level was –1134.0 µm/m with a standard deviation of 0.3 µm/m (equals 0.03%) for the three repetitions. This compressive strain increased by approximately –113.5 µm/m per force increment, and the maximum standard deviation of 0.9 µm/m was found at the penultimate load level. The measured strain levels for the holding time of 5 s for the individual force levels were stable with a maximum standard deviation of 0.1 µm/m. The percent deviation compared to the simulated strain was <0.68% for each force level (apart from the initial situation with 5 N preload). At the maximum force level, the deviation amounted to 4.3 µm/m (equals 0.38%). The good accordance was also reflected by the linear regression between the experimental and measured data as depicted in
Figure 7a, with a slope of 0.9932 and a coefficient of determination R² of 1.0000.
The vertical displacement of the actuator corresponded to the displacement of point
C in vertical direction
. The average measured maximum value at the maximum force level amounted to 0.74 mm with a standard deviation of 0.0019 mm (equals 0.26%), which was also the maximum standard deviation for all force levels after the preload. The measured displacements
for the holding time of 5 s for the individual force levels had no large fluctuations with a maximum standard deviation of 0.0005 mm; it was always below 0.3% of the mean value. Regarding the simulated displacements
, the experimental values for each force level were approximately 1.5 times higher (see
Figure 7b). Therefore, the linear regression line had a slope of 1.5713. The coefficient of determination (R² = 0.9998) was comparably high to the coefficient of determination for the strain
. The absolute maximum deviation occurred for the highest force level with 0.2598 mm (53.76% compared to the simulated displacement
).
The loads measured by the load cell for the different force levels were in accordance with the input data. Deviations were <0.003% for the averaged forces at all levels for the three repetitions. The standard deviation at the individual force levels for the holding time of 5 s per level was always <0.15 N. The good accordance of the force input and the measured force also for the dynamic testing is reflected by the graph in
Figure 7c. The percent deviation between the absolute maximum and minimum for an arbitrary cycle was <0.6% and <0.3%, respectively.
Figure 8 displays the simulated voltage for the stacked piezoelectric multilayer element, assuming the nominal capacity from the manufacturer with the given tolerance and the calculated capacity according to Safaei et al. [
56] for the load resistance of 0.5 MΩ (maximum power output for experimental testing, see below). It was compared to the data for the experimental testing of the piezoelectric element only. These results were gathered by directly loading the piezoelectric element with the force profile, which was also used for the numerical calculation (also shown in
Figure 8).
The voltage rose with the increasing force. When the load slope flattened and the force approached the first local maximum, the voltage dropped. It rose again with the second force peak. The unloading to the force minimum drove the voltage to negative values. This pattern was repeated for all cycles, with the first voltage maximum more pronounced in the initial cycle.
The two calculated voltage curves were in very good accordance regarding their course. However, the approximated capacity led to smaller absolute voltage values than for the nominal capacity from the manufacturer (e.g., the first maximum decreased by 22%). The difference between the experimental data and the data calculated on the basis of the nominal capacity from the manufacturer was even more pronounced (e.g., the first maximum decreased by 37%), but the trend matched very well.
For the implanted system, the measured voltage at a load resistance of 0.5 MΩ is shown in
Figure 9, together with the data for the directly loaded piezoelectric element (already known from
Figure 8). The curve progressions were in very good accordance and the absolute values of the generated voltage for the implanted system were notably smaller. For example, for the first peak, the maximum was reduced by 46%.
The generated power for different load resistances, calculated from the different voltage curves, is shown in
Figure 10. The calculated power maximum amounted to 72.7 µW at 0.68 MΩ, assuming the nominal capacity from the manufacturer. When approximating the capacity according to Safaei et al. [
56], the maximum was decreased by 35% to 47.4 µW and the corresponding resistance shifted to 0.44 MΩ.
The maximum power output for the experiment amounted to 28.6 µW for the directly loaded piezoelectric element and 10.2 µW for the implanted system, both for a load resistance of 0.5 MΩ. Regarding the data calculated on the basis of the nominal capacity from the manufacturer, this equalled a reduction of 61% and 86%, respectively. Between the directly loaded piezoelectric element and the implanted system, the difference was 18.4 µW (equals 65%). The curve progressions were all in accordance, showing a rise to the maximum power, followed by a slow decrease towards higher resistance.