Thick Film Sensor Manufacturing Techniques for Realization of Smart Components via Low Pressure Die Casting

Recent developments like autonomous driving have increased the interest of the automotive industry in structural health and condition monitoring of safety-relevant components often manufactured via High or Low Pressure Die Casting (HPDC, LPDC). Specifically the latter process is important in this field as the improved quality of castings produced facilitates T6 heat treatments aimed at optimizing mechanical performance of such components. The specific challenge associated with this approach is that it produces critical thermal loads which may compromise the characteristics of the integrated sensor. The present study shows the results obtained by LPDC on thick film sensor systems prepared by means of screen printing on aluminum substrates. The response of the sensor systems to the thermal loads associated with the casting process is evaluated in casting experiments, with temperatures reached during casting checked by means of thermocouples attached to the inserts. The focus is on the influence on general properties of the piezoresistive sensors as well as on their response to thermal and mechanical stimuli. The results show that in contrast to HPDC, in the case of LPDC, the former provides stimuli for thermally induced processes which can compromise sensor layers. The respective trends have been studied via resistance measurements on samples subjected to repeated firing cycles and actual casting experiments.


Introduction
An increasing number of sensors is integrated in cars every year.New trends such as autonomous driving create more necessity for real-time monitoring of the status of the car and its surroundings [1].Integrating sensors directly into car parts during manufacturing could yield highly precise data of the actual state of the car's main structural elements, and in particular of safety-relevant chassis components [2].This would allow to sense directly at the point of origin of potential damage, deep inside the metal part [3] [4].However, the harsh processing conditions, which, depending on casting process and material, entail exposure to high temperatures and mechanical loads.A broad overview of the respective research area has recently been published by Lehmhus et al. [5].
Related studies on integration of thick film sensors in high pressure die casting (HPDC) of aluminum alloys have previously been performed by our working groups [3] [4] [6].In this process, roughly ¾ of the sensors survived the casting process.Mechanical evaluation of the respective sensors in integrated state by means of cyclic three point bending tests showed very good correlation between the displacement signals gathered from the universal testing device and the data recorded by the integrated sensors.Moreover, it was possible to identify failure initiation based on the sensor data [3] [4] [6].
The aim of the present study is to also evaluate the sensor system already proven in HPDC as well as some derivatives in low pressure die casting (LPDC).The challenges in this respect are the higher thermal loads in LPDC, as well as the complication added by the fact that these castings are usually subjected to precipitation hardening, consisting of a solution heat treatment (SHT) at temperatures slightly below the solidus temperature of AlSi-based casting alloys, followed by either natural or warm ageing.For the common aluminum LPDC alloys based on the Al-Si eutectic system, the first of these two steps typically means heating the castings to temperatures of approximately 520-540°C and holding them at this level for several hours, followed by quenching.In contrast, while temperatures endured by the sensors during casting are somewhat higher (up to and above 600°C), the time above, say, 400°C, is measured in no more than seconds for HPDC and still just minutes for LPDC, making SHT the critical process step, as it provides sufficient energy as well as ample time for diffusion processes to occur and damage the sensor system.
However, it is a commonplace that structural health monitoring (SHM) is justified primarily for safety relevant parts.From an automotive industry perspective, these are mainly chassis components produced by LPDC.The prerequisite for endowing such parts with self-monitoring capabilities are thus sensors which can sustain the conditions implied by the casting process.This is typically achieved using thick film insulation to protect the sensor [7] [8].The sensor themselves must also survive the high temperatures and stresses linked to casting and cooling [9] [10].Under such conditions, insufficient matching e.g. of coefficients of thermal expansion (CTE) could lead to failure due to thermal stresses [11] [12] [13].The present study addresses the aforementioned challenges via a thick film system based on a set of functional materials tailored for compatibility with aluminum alloys.As substrate, the wrought aluminum alloy AlMg3 (EN AW-5754) is used, while the casting alloy is AlSi7Mg0.3.This ensures that substrate and cast material have sufficiently similar mechanical and thermal properties, while AlMg3 brings in a comparatively high solidus temperature.

Sensor Fabrication
In low pressure die casting, the complete sensor must endure high temperatures and thermal stresses during casting, in the course of which it is engulfed by molten aluminum reaching temperatures between 620 and 730 °C [5].Therefore, it is manufactured on top of a 1.5 mm thick AlMg3 sheet (160 x 30 mm) and fully encapsulated.The selection of AlMg3 (EN AW-5754) is based on this alloy's comparatively high melting point, with approximate solidus and liquidus temperatures at 595°C and 645°C, respectively (determined using JMatPro ® for the average nominal composition according to [14]).To ensure proper insulation of the sensor, the screen-printable paste IP6080A (Heraeus Deutschland GmbH & Co. KG) is used.This product is a dielectric material specifically developed for use on aluminum substrates, to which it is matched e.g. in terms of properties like the coefficient of thermal expansion.Six consecutive insulation layers of IP6080A were screen-printed and fired at a peak temperature of 570°C to avoid electrical connections or leakage between substrate and sensor layer.Afterwards, a resistive carbon-based paste was screen-printed on top of the insulation layers and fired using the same parameters.This paste (PCR12020 Heraeus Deutschland GmbH & Co. KG) is compatible with the IP6080 paste and thus ensures proper adhesion to the latter.Covering the sensor layers, 5 additional layers of IP6080A were printed to provide insulation against the cast material.The piezoresistive sensor features a meander shape to multiply the effect of the applied strain.Four sensor structures were produced for each aluminum substrate.The dimension of sensors were chosen to be as small as possible, while still maintaining the electrical connection.Due to the feature size of the applied screen-printing mask (SEFAR PET 1500, 50μm openings), the minimum line width was chosen to be 200μm, with a spacing of 250μm.A schematic of the sensor layout is shown in Fig. 1.

Low Pressure Die Casting (LPDC) Experiments
LPDC casting experiments were performed at Fraunhofer IFAM's facilities at the Open Hybrid Lab Factory (OHLF) in Wolfsburg on the TEGISA II low pressure die casting machine, which is characterized by a melting power of 130 kW to be used in direct and indirect inductive heating, a maximum temperature of 1650°C and a casting pressure of up to 1.0 bar The melt temperature was set to 750 °C, which is common for casting an AlSi7Mg0.3(AC70 DV) alloy.The actual temperature in the furnace was recorded prior to each casting cycle, exhibiting a range between 742 and 758°C.The die used in the experiments is of horseshoe geometry with 5 insert positions in each arm, allowing the integration of inserts of up to 2 mm thickness in one arm and up to 4 mm thickness in the other.The positions are numbered from bottom to top, with the lowest position thus corresponding to the section exposed to highest melt temperatures, and extended exposure times (see Fig. 2).During the experiments, the positions number 3 and 4 on the 2 mm side were used to evaluate the effect of different levels of thermal load on the sensors.In order to achieve homogeneous mold filling, opposite to the insert on the die's 2 mm side, for each experiment, a dummy insert was placed on the 4 mm side.Figure 2 depicts the die in opened state with the casting standing upright in between the two die halves.Die temperature was set to a nominal value of 350°C.Surface temperature measurements on either die half adjacent to the insert position showed a range of actual temperatures from 227 to 321°C.This is due to the progression towards a steady state of the die as casting continues.Superimposed are the effects of deviations in die open and general cycle time caused by the experimental character of the casting process.Given the thickness of the casting (15 mm), it is safe to assume, though, that the temperature of the interface between insert and cast material is mostly dominated by melt temperature and heat transfer between melt and insert.For an insert in the central position 3, in situ temperature measurements using embedded thermocouples in the insert's central position confirmed maximum temperature level of approx.608-615°C for insert positions #2 (see Fig. 2, counting from below with position #1 being the one nearest to the ingate), #3 and #4.

Sensor Resistance Measurement Setup
The influence of the firing process on the sensing layer was characterized by measuring the sensor resistance after each firing cycle with a digital multimeter (40 MΩ range).It was expected that the resistance of the sensors would change due to the repeated exposure to high temperatures during consecutive sintering of the top insulation layers.Similar measurements were conducted after the casting process.Sensor performance in terms of the response to mechanical loads was determined via a forcecontrolled, symmetric three-point bending test performed at a force amplitude of 50 N and a frequency of 0.2 Hz prior to casting (DYNA-MESS 1720DLL-10KN).The distance between lower supports was 140 mm.A commercial strain sensor (HBM 350Ω, gauge factor 2.1) was glued onto the substrate to serve as reference.This reference sensor was positioned directly adjacent to the sensor's meander structure, to ensure that the same strain was acting on both reference and printed sensor.

Sensor Structuring
Each insulation layer has a thickness of ~20 μm, resulting in a total thickness of the top and bottom insulation of 100 and 120 μm.An example of the sensor structure on the bottom insulation layer (without top insulation) is shown in Fig. 3a.A closer view of the connection track to the meander structure is shown in Fig. 3b.The high viscosity paste is screen-printed through the mesh and fired at the recommended temperature, producing a sensing layer with a thickness of ~6 μm.To ensure a maximum achievable smoothness of the surface, after each screen-printed layer (insulation and sensing layers), the sample is placed to rest on a flat surface for at least 30 minutes.This enables the viscous paste to level, minimizing the layer roughness.However, after inspecting the layers, limited topography variation remains visible corresponding to the mesh openings (visible in Fig. 3a).

Effect of Firing on Piezoresistive Layer Properties.
The electrical resistance of the samples is affected by the high firing temperatures.Adhesion between the sensing and insulating layers, and also of the latter to the substrate, was investigated by performing a so-called quench test: To this end, a set of samples were fabricated without top insulation, heated to 550°C and placed in cold water immediately afterward.The measurements indicated that the resistance of sensors increased by ~10% after quenching.The layers themselves did not show any visible surface modifications.All layers remained intact and adherent both to each other and the substrate.Slight increases in resistance were also observed after each firing step of the top insulation layers, as shown in Fig. 4a.Average resistance in pristine state was 22 kΩ.A small spread in resistance had been expected due to limited thickness and feature size variations commonly observed in screen-printing.After firing of the first top insulation, a most sensors showed a sharp increase in resistance reaching almost 100%.Subsequent firing steps resulted in a linear trend up to a level of approx.65 kΩ after the 5 th firing.Individual sensors showed a rise in resistance up to the level of an open circuit (>40 MΩ), which indicates damage introduced during firing: It is assumed that small impurities introduced during the screen-printing process caused small gaps in the thin sensor lines that grew during the following firing steps.For sensor fabrication, 24 aluminum substrates were used, having each of them four screen-printed sensors.Of the 96 sensors, 91 were in working condition before top insulation layers were added.After application of the latter, ~87% had survived the complete process prior to casting.A detailed overview of sensor failure per top-insulation firing step is shown in Fig. 4b.

Sensor Response to Mechanical Load
The response of the sensors to mechanical strain was analyzed by means of cyclic three-point bending test.The samples were placed face-down to protect the sensor layer during the test.The force amplitude was set to a maximum value of 50 N to avoid any damage of the sensor prior to the casting experiments.This configuration results in a tensile strain of ~0.1%, which was determined via the reference sensor.The commercial sensor and the sample were connected via a multiplexer to a digital multimeter to capture all data.Measurements confirmed that the slight deformation did not cause any noticeable damage or permanent deformation to the samples.The test frequency was set to 0.2 Hz.Altogether 100 load cycles were conducted.During testing, the change in resistance of the screen-printed sensor matched the results obtained with the commercial reference sensor, as can be seen in Fig. 5a.The printed sensors show an initial drift towards a lower resistance change until stabilizing at roughly 40 cycles (see Fig. 5b) and subsequently exhibiting a stable response during the remaining cycles.The measured resistance values thus provide repeatable data points without sudden deviations.It is reasoned that the test deformation is well within the working region of the sensor.

Effect of the Casting Process on Sensor Characteristics
Twenty aluminum substrates (74 working sensors) were embedded in an LPDC process.The substrate carrying the sensor is protruding from the casting, with conductive paths and contact pads clearly visible (Fig. 2b), while the actual sensors are located within the casting.The resistance of the sensors was measured once again after casting and, as expected, the upward trend continued (Fig. 6a).The fact that the level of increase exceeded that observed during the consecutive firing steps is linked to the higher temperature reached during casting.Of the four sensors printed on each substrate, three failed in almost all cases.However, of one specific sensor on each substrate, a total of 70% survived the casting process.It is thus supposed that the sensors as such can indeed survive the casting process, though in practice, defects introduced during sensor fabrication caused some to fail.This was later confirmed by an examination of the screens utilized for resistive layer deposition.Light microscopy images (not shown here) showed that the mesh openings in three out of four sensor structures were clogged with dried paste or had not been completely opened during the structuring of the masks.This resulted in small gaps in the conductive lines (Fig. 6b), which likely caused the critical failure of three sensor designs during casting.

Conclusion and Outlook
Integration of printed sensors in LPDC process chains is a difficult task, as thermal loads are considerably higher than in HPDC, for which the applicability of the approach had previously been proven.The current study has investigated stability and performance of comparable systems produced via thick film techniques, notably screen printing, using commercial, carbon-based pastes for realization of the piezoresistive sensors, along the sensor production process chain and up to and including the casting step.
The capability of a high percentage (75%) of the sensors to survive the casting process was proven.Loss of functionality of additional sensors could be linked to defects introduced during sensor production rather than the casting process, a source of failure which can be remedied via printing process and sensor design adaptations.Sensor resistance was observed to rise during each individual firing step of the top insulation layers required to electrically separate sensor and cast material.The increase of resistance measured amounted to roughly 32% across all five steps, to which casting added a further 16%.However, the sensor function as such was not affected by this increase, as could be demonstrated by mechanical tests (3-point bending) performed on sensors following the firing of the last top insulation layer, which showed very good agreement with the reference data provided by a commercial strain gauge attached adjacent to the printed sensors and read out in parallel to them during testing.
Future research will be focused on increasing the survival rate, evaluating the effect of the typical T6 heat treatment applied to automotive components as addressed by the present project (for this alloy, solution heat treatment at 535°C for several hours as critical step) on this type of sensor, and characterizing casting-integrated sensors in terms of thermal and mechanical sensitivity.Future research will also have to focus on the question of the mechanical interaction between the sensor system, including its substrate, and the casting.Essentially, this relates to the field of compound casting, i.e., the approach of using the casting process to create a joint between an insert and the cast part itself.For a structural health monitoring context, this is an important issue in two respects: For one thing, if strains are to be measured, it is essential that these are transmitted from the monitored component to the sensor.In the present case, this must either be achieved via the sensor substrate, or via the top insulation.Since initial, qualitative evaluations show that no bond is formed between insulation and cast material, mechanical coupling must rely on the substrate.Besides, in case of insufficient bonding between insert and casting, the former may act as a wound within the component, and thus limit its performance [15] [16].While one potential remedy is the reduction of sensor and thus insert size as proposed earlier by Lang et al. [15], supporting the formation of a tight bond is another: The latter approach has been discussed in several studies on Al-Al compound cast components [17].However, a direct transfer to sensor integration is still pending.

Figure 1 .
Figure 1.Schematic representation of the sensor layers.Note that the relative thickness of sensor build-up and substrate (insert) deviates from the real ratio in this generalized image.

Figure 2 .
Figure 2. Die and casting with sensor in position #3.Metal sheet visible in position five is a dummy insert used to close the gap.These are used in all positions, but usually fall off.

Figure 3 .
Microscopic view of the sensor structure: (a) Overview of the sensor layout with orthogonal traces of the printing mesh visible on the bright insulation layer adjacent to the dark conductive paths, (b) detail view of the latter.

7th
International Conference of Engineering Against Failure Journal of Physics: Conference Series 2692 (2024) 012007

Figure 4 .
Influence of the sequential firing of the top insulation layers: a) Change in electrical resistance of sensors after each top insulation layer.b) Total sensor survival rate after each top insulation firing step before casting.

Figure 5 .
Typical sensor response of screen-printed sensors: a) Comparison with commercial sensor output.b) Stability test for 100 cycles.

Figure 6 .
Effect of the casting process on thick-film samples: a) Resistance increase of a typical sensor up until after casting.b) Fabrication failure due to clogged mesh openings in the screenprinted structure.