1. Introduction
Aerospace components demand lightness as an essential design characteristic, a particularly relevant requirement in the space industry as well. Indeed, mass reduction is considered one of the most valuable drivers during the thermomechanical design of space hardware and aerospace components [
1,
2,
3,
4]. This objective can be effectively tackled by coupling the choice of high strength-to-density ratio materials with structural optimization, resulting in performant and lightweight components. However, due to their excessively high costs in procurement and manufacturing, qualification, and certification, [
5] innovative materials have had low mileage concerning the design of spacecrafts components. Conversely, structural optimization tools played a significant role in the aerospace industry. Optimal design proposals can be obtained by exploiting a streamlined design-to-manufacturing track.
Among the non-conventional thermomechanical design approaches, the Topology Optimization (TO) perfectly fits this framework. The TO couples mathematical optimization algorithms to Finite Elements Analysis (FEA) to suggest structural design outcomes ensuring an optimal material distribution within a given reference domain. Specified optimization objectives, design constraints, loads, and boundary conditions are applied to the study case [
6]. Stresses, displacements, natural frequencies, and geometrical features are typically encompassed as design constraints towards the achievement of an optimal solution, although they can also be included as the TO objectives to be pursued within the optimization iterations [
7].
Since the optimal material distribution may lead to complex geometry, the TO outcomes can become ready-made Additive Manufacturing (AM) designs. Alternatively, a further post-processing stage can be introduced to adapt the design for traditional—material subtractive—manufacturing techniques. Nevertheless, this strategy allows to bypass the time-consuming engineering and iterative design mechanisms, as the optimal solution coincides with the outcome of the process itself. The TO is rapidly widespread in aerospace components design and improvement, and a multitude of examples can be retrieved in the literature [
8,
9]. Worth mentioning are the VEGA launcher connector support [
10], a structural component obtained via a frequency-driven TO, or the Airbus A320 nacelle hinge bracket [
11], an aircraft part manufactured exploiting Additive Layer Manufacturing (ALM) after its optimization. Another significant example is the antenna support for the ESA Sentinel 1 satellite [
12,
13], a component redesigned exploiting TO towards the achievement of mass minimization and finally manufactured via Selective Laser Sintering (SLS).
This research work extensively describes the MicroMED Dust Analyzer Optical Bench (OB) TO-driven redesign workflow and results assessment activities. The instrument is an amplitude-dependent single Optical Particle Counter (OPC) designed and qualified for the ESA ExoMars 2022 mission towards Mars [
14,
15,
16]. The instrument includes several subassemblies and hardware components arranged in a compact and lightweight design, whose locations are depicted in
Figure 1 and was intended to infer the airborne dust size and size distribution in situ at the lowest troposphere of the Red Planet [
17,
18,
19]. This would be a groundbreaking achievement because, at present, data about dust on Mars were exclusively acquired using remote sensing (e.g., thermal IR instruments, near-IR spectrometers, and Terrestrial ground-based telescopes).
A proper fluidic system was designed to suction the Martian aerosol—a multiphase mixture of gases and solid dust particles—and transfer it to the MicroMED Optical Head (OH). There, the fines are invested by a collimated low-power NIR laser beam (having a single-mode nominal optical power of 150 mW and a lasing wavelength of 830 ± 5 nm) and finally returned to the Martian environment. The scattered light produced by the interaction between the particles and the laser is conveyed by a parabolic mirror towards a photodiode detector, responsible for the particle size measurements. MicroMED functional components and subassemblies are mounted on the OB and shielded from the harsh Martian land by the External Envelope (EE) [
20,
21,
22].
The activities described in the following aim to complete the redesign work aiming to obtain two different optimal OB design proposals with an improved performance. The feasibility design was successfully carried out by the author in [
23], but the numerical investigation has been extended, focusing on the evaluation of the thermo-elastic behaviour of the proposed solutions in the worst-case temperature. Moreover, the experimental assessment of the modal behaviour of a mockup of one of the two optimal solutions is presented, aiming to validate the presented methodology and related results.
The added value of this research work lies in the innovative approach adopted for the OB improvement. The latter is presented in
Section 2 along with the FEA procedure involved within the parts’ numerical assessment, whose results are summarized in
Section 3. The achievements consist of optimal design proposals to improve the current—suboptimal—material arrangement within the OB. Starting from the achieved design proposals, as a further improvement of the research activity a mockup of the lightened OB was manufactured and tested to prove the effectiveness of the TO design. Finally, the experimental validation campaign is presented in
Section 4, comparing the outcomes with the numerical ones and completing the OB optimization activity.
2. Materials and Methods
The MicroMED OB is the support plate on which the instrument subassemblies and functional components are firmly fixed. The OB is a ribbed aluminum component (i.e., manufactured in Al7075-T6 aluminum alloy) characterized by high mechanical stiffness, provided by the grid of ribs lying on its interface plate. In this research work, the main investigation dealt with the topology and sizing of the OB ribs, pursuing the optimal configuration towards mass minimization achievement (i.e., hereafter referred to as Model A) maintaining comparable dynamic performance with respect to the current OB design. Furthermore, a different layout was probed, aiming to optimize the OB dynamic performance (i.e., hereafter referred to as Model B). The latter proposal was obtained by maintaining the unchanged mass budget allocated to the ribs grid while maximizing the OB resonance frequencies.
The TO method was preferred to the other structural optimization approaches, i.e., the size and the shape optimization methods, because of its design freedom. Both Model A and Model B redesign studies exploited two commercial software programs, i.e., PTC Creo Parametric/Simulate and Altair Inspire. The former was adopted for computer-aided modelling and FEAs, while the latter was employed for the TO. A schematic of the workflow exploited for the redesign is depicted in
Figure 2.
Within the optimization process, Altair Inspire relies on a computational method based on the material distribution technique. This process adopts the material density as the design variable and returns an optimal material distribution in a discretized reference domain [
24,
25]. A binary solid model is produced, and for each reference domain finite element either a unitary or null value is assigned to the design variable. Consequently, a finite element will be finally part of the optimal structure only if its associated design variable has a non-null value. In the latter case, the material density is assigned to the finite element. The software engine, to cope with the occurrence of the intermediate values of the design value (i.e., between 0 and 1) during the optimization iterations, adopts a penalization approach based on the Solid Isotropic Material with Penalization (SIMP) algorithm [
26,
27]. Therefore, the outcome will be composed of either voids or solids.
In this research work, the optimization cycles imposed the finite element structure mechanical stiffness maximization as the design target. At the beginning of each iteration, the overall mass of the reference domain was progressively reduced by a percentage within the total reference domain volume. For the OB optimization, at the beginning of each TO step, 70% of the design volume removed was imposed as an output condition. Before running the optimizations, a filled OB solid model was developed, as depicted in
Figure 3. Solid model defeaturing was carried out as the first task to reduce the computational complexity of the FEAs, while maintaining the representativeness of the numerical outcomes.
The Finite Element Model (FEM) held three lumped masses connected via rigid links to the OB areas intended to support the instrument subassemblies. Such areas were excluded from the design domain and thus were not involved in the optimization cycles. These features—92 g for the OH + Inlet/Outlet (IO), 51 g for the Pump Group (PG), and 29 g for the Laser Group (LG)—were introduced to consider the main instrument’s subassemblies but in the meantime to run a streamlined optimization process. The lumped masses were estimated by assigning the aluminum density (i.e., 2819 kg/m
3) and assuming a 10% margin as the worst-case scenario. The material depletion of the design domain (
Figure 3), i.e., the only region eligible for the TO, was performed by generating constant height ribs in the direction perpendicular to the OB floor. No voids nor holes throughout the height of the ribs were allowed for the final design.
The loading condition applied for the optimization was a quasi-static acceleration of 1000 m/s
2 on each lumped mass, in each direction. Each iteration for Model A concluded only when the result produced a new model having the first natural frequency fulfilling the design dynamic requirement, thus when the first three resonances were equal or larger than the current OB ones. On the other hand, for the Model B development, the optimization was interrupted only when the optimal material distribution was identified, i.e., the one providing the highest value possible for the component’s natural frequencies. In this case, the design constraint was set on the mass budget spent on the ribs grid, imposed equal to the current OB one. Moreover, as pointed out in
Figure 2, following the TO, a post-processing activity was performed on the optimal solutions, with the aim of smoothing and homogenizing the optimal ribs grid towards a milling/Electrical Discharge Machining (EDM) manufacturing process. Preliminarily, a FEM of the current OB was developed in PTC Creo Parametric/Simulate environment to define the baseline performance to be exploited during the TO. As shown in
Figure 4, pin constraints were placed at the OB peripheral holes to simulate the MicroMED instrument mounting interface on the Dust Suite, the ESA ExoMars 2022 lander surface platform. Mesh refinement was applied to the FEM, obtaining an overall of 32,371 finite elements, split between 28,704 solid tetrahedrons and 3667 shell triangles. Likewise, the FEMs for the optimal and post-processed OBs were developed. The performance of the models was estimated relying on the FEAs set:
Modal analysis aiming to estimate the first five natural frequencies;
Quasi-static analyses applying a load of 1000 m/s
2, probing the state of stress distribution of the models considering three different loading directions (i.e., X, Y, and Z according to the reference frame reported in
Figure 4) and resembling the excitation expected during the takeoff and landing phases of the mission; and
Thermoelastic analyses in either non-operational (i.e., the cold case, considering a temperature shift from the reference +20 °C to the target −40 °C) or operational (i.e., the hot case, likewise the cold case but within the range +20 °C and +40 °C) conditions.
The figure of merit to state the quasi-static and thermoelastic FEAs goodness was the Margin of Safety (
). The
was evaluated to highlight the resistance of the mechanical component during the assessments, relying on the Von Mises yield criterion for ductile materials. The
was computed referring to the following formulation:
In the equation,
was the material yielding stress (i.e., 503 MPa for the Al7075-T6),
was the admissible safety factor (i.e., 1.5 as reported in the ECSS-E-ST-33-01C design standard [
28]), and
was the maximum Von Mises stress resulting from the simulation. The assessment was considered satisfactory with positive
.
Moreover, the thermal deformations induced within the temperature ranges were probed both for the current and the optimal OBs. The intent was to quantify the impact of the temperature on the structural stability of the MicroMED hardware, i.e., estimating the misalignment between the laser light and the aerosol flux due to the thermal deformations of the mechanical parts for the cold and hot cases. As depicted in
Figure 5, the OH was assembled with the OB; on the former, the Laser Delivery Group (LDG, composed of the 213 and 11B parts) was fastened, as well as the IO, which was split between the Inlet Tube (IT, composed by the 022 and 026 parts) and the Exhaust Tube (ET, or the 023 part). A cloud of probes to measure the displacements along the X, Y, and Z axes was introduced at the LDG, IT, and ET extremities facing the sampling volume inside the OH. The relative shift among the probes allowed for an estimation of the misalignment at the instrument sampling volume. According to its magnitude, the latter can produce a negative impact on the MicroMED performance, jeopardizing the particle size identification and increasing the overall measurement uncertainty.
4. Experimental Assessment
As mentioned before, mass minimization plays a key role in aerospace components design. In this light, among the proposals probed within this research work, Model A was considered the most promising one for MicroMED mechanical hardware future development. Despite the optimal OB robustness being positively proven numerically via FEAs, for completing the circle an experimental activity to further freeze the adopted workflow was carried out. Indeed, a mockup of the Model A was manufactured by milling from an Al7075-T6 slab (
Figure 16) and was tested to measure the free natural frequencies. Before the experimental testing activities, a dimensional and geometrical check was performed to state the actual mockup footprint, to update the FEM.
The mass was measured with a Gibertini balance having a resolution of 0.01 g, obtaining 63.88 g. The mass mismatch with the Model A Computer-Aided Design (CAD) was attributed to a slight difference between the real and theoretical component dimensions. Indeed, a measure of the ribs grid and the overall OB geometrical features—carried out with a Mitutoyo fiftieths gauge and a Micro-Epsilon laser scanner having a resolution of 2 µm—confirmed the presence of a dimensional unevenness. The measurement outcome is reported in
Appendix A, providing a comparison between the theoretical and real dimensions of Model A.
The activity was propaedeutic to the actual free vibrations test because it allowed updating the Model A CAD model with the real mockup dimensions and preserving its representativeness. This was mandatory in view of a comparison between the numerical free natural frequencies (i.e., found running a free modal analysis on PTC Creo Parametric/Simulate) with the experimental one.
Moreover, as highlighted in
Figure 16 during the manufacturing process the 0.3 mm skin was not able to withstand the milling head-cutting force and was plastically deformed by the tool action. However, since the damage was localized to a restricted area of the component, the mockup was considered eligible to be investigated in this experimental activity. Indeed, its structural integrity was not compromised and the vibrating modes not involving the damaged region were expected to be unaffected by the manufacturing defect.
Finally, before the actual resonances research experimental activity, the Model A updated CAD was subject to a free modal analysis to estimate the free natural frequencies and discover which vibrating modes were eligible to be observed during the resonances research. Indeed, considering a non-contact measurement method including a Laser Doppler Vibrometer (LDV—Polytec OFV 505 sensor head) was used for the experimental activity, the free modal analysis was necessary to plan where to point the laser head to properly catch the vibrating modes and measure their frequencies.
The FEA outcomes are reported in
Figure 17 and the numerical results are summarized in
Table 4. As can be seen, with only three LDV laser spot positionings (i.e., at one of the peripheral mockup holes, at the skin between the LG and the PG, and near the component damaged area) it was theoretically possible to catch all the investigated vibrating modes.
An overview of the adopted experimental setup is depicted in
Figure 18. For the free vibrations test, the Model A mockup was suspended on elastic ropes and excited with a series of sweep sines (i.e., five consecutive linear sweeps lasting 30 s each, in the 80 to 5120 Hz frequency range and 200 mV peak-to-peak wide) produced by the Keysight Technologies 33220A function generator and delivered to a loudspeaker placed below the OB. The real input delivered at the mockup was measured with the PCB Piezotronics 130E20 ICP microphone (i.e., having a sensitivity of 48.1 mV/Pa). On the other hand, the mockup response was measured relying on the LDV (i.e., having a sensitivity of 40 mV/(mm/s) and a band-pass filter between 100 Hz and 10 kHz) and sampled at 10,240 Hz via the National Instrument NI 9234 acquisition board (i.e., characterized by an input range of ±5 V and 24 bit). The LDV beam was pointed at three strategic positions within the Model A mockup according to the free natural frequencies mode shapes identified via the free modal analysis. Finally, the experimental free natural frequencies were estimated relying on the Frequency Response Function (FRF) computed with the ratio between the LDV and the microphone signals, windowing (Hanning) and averaging the acquired time histories. Results are shown in
Figure 19.
A comparison between numerical and experimental natural frequency values is reported in
Table 5. As can be seen, all the investigated resonances—but the 5th—are in line with the numerical ones, i.e., the retrieved error is smaller than 5%. The discrepancy between the numerical and the experimental natural frequency value at the 5th vibrating mode was explained considering observing the mode shape reported in
Figure 17. Indeed, the considered mode exclusively involved the thin skin affected by the manufacturing damage pointed out in
Figure 16. Therefore, such local weakening may have sensibly reduced the component stiffness there and caused the natural frequency value to drop.
Nevertheless, in light of the satisfactory results with the comparison, the validation activity to state the robustness of the redesign approach was considered positive.
5. Conclusions
This research work proposed the redesign of the optical bench for the MicroMED Dust Analyzer, an OPC developed for the ESA ExoMars 2022 mission to Mars. Topology optimization was exploited to achieve either a mass saving on the component (i.e., Model A) or a maximization of its dynamic performance (i.e., Model B). The optimal proposals allowed a reduction of about 55% of the mass budget of the considered design domain, or a first natural frequency improvement of about 50% than the current optical bench design. Both of the solutions were positively assessed via a set of FEAs (i.e., modal, quasi-static, and thermoelastic analyses) applying the conditions identified as design constraints within the mission framework and highlighting large safety factors about the limit stress for both the foreseen mechanical and thermal environment. Moreover, the thermo-elastic analyses showed that computed thermal deformation is negligible with respect to the instrument’s measurement volume size, therefore validating the thermo-mechanical structural design.
The Model A was considered the most promising proposal for the MicroMED future development. Indeed, applying the mass reduction to the overall instrument is foreseen to provide the largest benefit to MicroMED. Any additional mass saving among the mechanical hardware would allow either the possibility of proposing a cost-competitive instrument, or a redistribution of the gained mass elsewhere to enhance the functionalities of the instrument. Considering the latter option, one very attractive feature would be the possibility to quickly and easily dismount or replace the instrument components even after the integration is completed.
Considering the attractiveness of the Model A proposal, a mockup of the optical bench was manufactured and tested to prove the robustness of the redesign workflow as well. An experimental setup was put in place to measure the mockup-free natural frequencies, obtaining a good agreement with the ones predicted via a numerical free modal analysis, i.e., a percentage error lower than 5% was found for the first four natural frequencies.
Nevertheless, both Model A and Model B proved the effectiveness of the adopted structural optimization approach. Thus, as this research activity develops in the future, the redesign methodology will be extended to the main mechanical components of the instrument.