Stress Corrosion Cracking of Tunnel Ventilation Fan Blades: A Case Study

Abstract

Installations in road tunnels, such as jet fans, are exposed to a harsh environment. After two failures of rotor blades, made of aluminum cast, a failure analysis was initiated. The main goal of this study was to determine the causes, e.g., overloading, material defect, corrosion, and design deficiencies, and to define actions to avoid such events in the future. The fractographical analysis showed corrosion attack and a cleavage type of fracture. These findings pointed to two damage mechanisms acting in both cases: stress corrosion cracking due to hydrogen embrittlement, causing a brittle static failure. The failure analysis did not lead to a full understanding of the damage mechanisms: Service loads were far too low to explain the high stresses needed for a brittle static failure. A wedging effect of corrosion products in the gap between the blade and the mounting was proposed as the cause of high static stresses in the fractured cross section. Alternative materials as well as different corrosion surface protections were evaluated with respect to corrosion and corrosion fatigue resistance. Finally, an appropriate surface protection was proposed as the solution for inhibiting corrosion and therefore the risk of the reproduction of such failures.

1. Introduction

1.1. Tunnel Installations

Road tunnels are designed for a long service life, i.e., more than 20 to 25 years for mechanical and electrical installations and more than 80 years for civil engineering works. From a building owner’s point of view, it is expected that these infrastructures can be operated with a minimum of maintenance effort. These are challenging requirements not only for reinforced concrete components but also for the design of metallic constructions, such as load-carrying elements of wall panels and false ceilings, or mechanical and electrical installations, such as lighting, ventilation, and signage, including complementary installations, such as cable trays. The environment in a tunnel can be aggressive, especially in geographical regions with cold winters, such as Switzerland: the atmosphere in a tunnel is corrosion promoting because the climatic conditions vary with temperatures between −20 °C and +40 °C and a relative humidity of up to 100%. On top of this, emissions from the road surface, such as dew-salt, exhaust gases, and abrasion particles from tires and brakes, can attack the metallic components. The regulatory authorities therefore have defined standards and design codes for the design of safety critical items: For example, the Swiss Federal Roads Office (FEDRO) classifies the metallic load-bearing and structural elements in road tunnels with regard to the materials and corrosion protection systems to be used [1,2]. These design codes should allow the designer to choose an appropriate material and corrosion protection system without going through a time- and cost-intensive verification program. In the specific failure cases in this study, the inconsistent application of these rules led to the failure of such safety-critical items: In two highway tunnels in Switzerland, the rotor blades of a tunnel jet fan failed during normal operation of the tunnels.

In the course of these events, a failure analysis was initiated in order to assess the circumstances that led to these unexpected failures, to assess the suitability of the materials used, and to define measures for preventing similar events.

1.2. Description of the Failed Fans

Both fan blade fracture failures occurred in a so-called jet fan mounted in road tunnels in Switzerland; see Figure 1 (CANON EOS 60D/Auto-Focus, -mode and -exposure/Ultrasonic lenses EFS 17–55 mm). These fans are used to induce a longitudinal air flow in the traffic space in order to control pollution and relative humidity during normal operation as well as to manage smoke propagation in the case of a fire incident. In the tunnel in which the first failure occurred, 18 identical jet fans were installed (8 and 10 per tube). The second failure occurred in a tunnel equipped with 6 identical jet fans (single-tube tunnel). Furthermore, the ventilation system of this tunnel consists of two exhaust fans located in the middle of the tunnel and a series of extraction dampers. As pollutant levels are continuously decreasing, the ventilation system is only operated under special circumstances, such as exceptional pollution levels, fire incidents, and regular tests. Therefore, these jet fans are inactive most of the time. According to Swiss design codes, the jet fans must be fully functional during a 2 h period at a constant temperature of 250 °C. This fire rating has to comply with EN 12101-3 [3].

The jet fans consist of a center (active) part, two cylindrical silencers, and the mounting structure. The center part itself is a rotor with 10 blades, installed in a cylindrical housing and driven by an asynchronous motor at 1500 rpm. The fan blades are single aluminum sand-casted pieces. These blades are mounted at their root by two mounting blocks; see Figure 2 (CANON EOS 60D/Auto-Focus, -mode and -exposure/Ultrasonic Lens EFS 60 mm). The fans are designed to have a service life of at least 20 years. Maintenance runs are performed typically once a year.

1.3. Circumstances of the Event

In 2012 and 2015, two failures of similar type occurred in two different tunnels in Switzerland.

The first failure occurred when the fan was running due to air quality requirements. One blade detached from the rotor at its root, whereupon all other blades broke. The investigation into the incident, which was completed before the second incident occurred, concluded that the cause of the failure was excessive vibrations combined with an insufficiently rigid casing. This hypothesis could not be substantiated by the fractographical findings: there was no evidence of cyclic loads on the fracture surface. As the strong vibrations detected were outside the tolerable range, the investigations were terminated with the conclusion that the construction must be made stiffer in order to avoid such strong vibrations. After the second incident, however, it became clear that this conclusion was not correct: No more vibrations were observed during the second incident, so this hypothesis for the fracture of the rotor blade became invalid.

The second incident occurred when a blade of a similar jet fan broke, again in the area of the root. Again, a failure analysis was carried out: The construction of the housing and the connection had been strengthened in the meantime so that the entire jet fan was considerably stiffer. Furthermore, the jet fan was not running at the time of the failure. Therefore, in contrast to the first incident, no strong vibrations could be detected during the run. It was obvious that the cause of the failure was something other than excessive loads caused by vibration. The corresponding analysis and results are described next.

1.4. Examinations on Site of the 2015 Event

The failed jet fan of the event of 2015 as well as two other intact fans installed in the same tunnel were visually inspected for general condition, mechanical damage, and corrosion. The typical dimensions of the system are given in

Table 1. Dimensions of the jet fan of the 2015 failure case according to the manufacturer’s specifications.

Variable	Failure Case 2015
Rotor diameter D	1000 mm
Number of blades	10
Root minimum diameter, D1	44 mm
Root outer diameter, D2	58 mm
Radius of the root to platform transition, ρ	10 mm
Blade total weight	2.2 kg
Blade wing weight	1.9 kg
Center of gravity of blade wing, r	0.3 m

. The ventilators of this tunnel did, in a first view, only show minor corrosion in the visual inspection. The rotor and its blades were covered by an organic painting as corrosion protection. Such protection was not present in the rotor of the fan that failed in 2012. The general condition of the systems was normal, with some signs of corrosion mainly at the screws and on the blades of the rotor: The organic painting was in some areas damaged, and corrosion, in the form of blistering and flaking of the coating, was observed but without significant material loss.

To characterize the corrosion exposure of these systems, wiping samples were taken from undamaged fans. The laboratory corrosion analysis consisted of the identification of anions and their concentration in an aqueous extract by capillary electrophoresis (Lumex 105, Lumex instruments, Fraserview, BC, Canada) using aqueous wipe samples from the blades’ surfaces. This analysis showed a high magnitude of chloride deposits of approximately 50–200 μg/cm². The corrosive properties of chlorides toward various metals were determined using the method described in [4]. In the presence of chloride deposits >10 μg/cm², a corrosion attack of metallic materials must be expected in the presence of a relative humidity of >50%. Because the relative humidity in such tunnels is most of the time higher than 50%, the environment is harsh.

2. Test Plan

The analysis of the rotor blade failures discussed here started after on-site inspection and documentation of the circumstances. The failed parts could be identified easily and were brought to the laboratory. The corrosion attack observed on the blades could easily be attributed to the aggressive environment and the lack of corrosion protection on the top surface. Even an initial visual inspection of the blades showed clear signs of corrosion. Therefore, the test plan defined covered not only mechanical but also electrochemical characterizations of the blade material:

• Visual inspection of the interfaces between the mounting and the blades;
• Corrosion chemistry analysis of the corrosion products found;
• Determination of chemical composition;
• Determination of material properties: static strength, impact fracture energy, fatigue and corrosion fatigue data;
• Fractography of the fracture surface;
• Metallography;
• Geometrical characterization and tip clearance measurement.

The main goal was to compare these data with the specified values to check whether the material of the defective parts met the specifications. In the course of the investigations, it turned out that the material fulfilled the specifications and the final failure was brittle fracture. Therefore, the next task after the laboratory results was to find a plausible cause for a high load that was high enough to cause a static fracture of the blade root. As will be explained in the coming sections, with the help of fracture mechanics, it was possible to describe a plausible and consistent scenario of these rotor blade failures.

3. Investigations on the Failed Part and on a Reference Part

The failed and an intact jet fan (as a reference) from the same tunnel were transferred to the lab for detailed examination. The failed part was disassembled, and examinations were performed.

3.1. Visual Inspection

The fractured section of the blade and the two mounting blocks were covered by white corrosion products; see Figure 1 and Figure 2. The surfaces of the fragments of the broken blade as well as all other intact blades were covered by an organic coating and showed clear corrosion phenomena in some areas: blistering and flaking of the organic coating and crusting of white aluminum corrosion products. The layer thicknesses of the organic coating of the blades were in the range of 50–70 μm (average). It can be assumed that the coating was applied to the fully assembled rotor and was more for visual aspects than for corrosion protection. The root of the blades, free of organic coating, was affected, especially at the throat and the platform opposite to the mounting, by strong corrosion phenomena: widespread corrosion leaving a voluminous crust of white corrosion products of aluminum and also a strong, localized corrosion attack (pitting corrosion); see Figure 2. In addition, the inner surfaces of the seats, free of organic coating, were affected by these corrosion phenomena.

3.2. Corrosion Chemistry

Three samples of white corrosion products, one from the failed blade and two from the reference blades, were taken for corrosion chemistry analyses with X-ray microanalysis using a SEM Hitachi S3700N scanning electron microscope with an Ametek OctanePro EDX detector (Ametek, Inc., Berwyn, PA, USA). The white products consisted mainly of aluminum and oxygen and between 1 wt% and 10 wt% of sodium, chlorine, and traces of silicon, magnesium, calcium, and sulfur. The solids were analyzed in a 1:20 aqueous extract. The concentrations of corrosive anions were determined using capillary electrophoresis: between 2.5% and 3.1% of chloride and up to 0.17% of nitrate were found.

3.3. Chemical Composition of the Aluminum Alloy

The chemical analysis of the blade material was performed using spark-ignition optical emission spectrometry (S-OES; Q4 Mobile, Bruker Co., Billerica, MA, USA). The mass distributions of alloy elements were measured on a sample from the failed blade and on two samples from a reference blade. The measurements were compared to the specification of the part, e.g., EN AC-48000 from DIN EN 1706 [5]. The composition showed a normal distribution of an AlSi12CuMgNi alloy: a silicon content of about 12 wt% and copper, nickel, and magnesium contents of about 1 wt%. This material is a precipitation-hardenable aluminum casting alloy with silicon, copper, and nickel as the main constituents, as well as manganese, magnesium, and titanium [6]. These alloys are suitable for various casting techniques, including sand casting. In comparison with other alloys, this material is heat resistant. Regarding the construction of fan blades, this allows for thin blades and therefore a reduction in weight and operational stress. However, the heavy metal content makes the alloy unsuitable for applications in extremely corrosive environments, but the corrosion resistance is sufficient for most technical purposes, including outdoor applications [7].

3.4. Mechanical Properties

From 6 blades, a total of 12 specimens were cut out from the core of the root to be tested in tension according to ISO 6892 [8] on a Zwick 200kN universal test machine (ZwickRoell GmbH, Ulm, Germany) and 18 specimens to be tested in a Charpy impact test according to ISO 148 [9] on a Losenhausen 150J Impact test machine (Maschinenbau AG, Düsseldorf, Germany).

The resulting yield strength (R_p0.2) of the cast aluminum alloy was between 115 MPa and 134 MPa, the tensile strength (R_m) was in the range of 186 MPa to 214 MPa, and the elongation at break (A_5.65) was between 0.8–1.5%. The tensile strength (R_m) was within the tolerance range for the material (EN AC-48000, T5 condition) according to DIN EN 1706:2010 [5]. The yield strength (R_p0.2) was partly below the minimum required value.

Impact fracture tests (KV150) showed fracture energies of only around 1 Joule, indicating low fracture toughness.

3.5. Macroscopic and Microscopic Fractography

The broken blade of the 2015 failure showed a rough and inhomogeneous fracture structure; see Figure 3a. In the outer areas of the blade root, the fracture surface was largely covered by a thick adherent crust of aluminum corrosion products (red-marked area in Figure 3b) as the core of the cross section was largely free of corrosion products. This aspect of the fracture surface indicated progressive corrosion-induced cracking at the outer surface with a purely mechanical final failure; see Figure 3c. After cleaning undamaged blades with dilute nitric acid, clear areas of corrosion in the form of pitting or crevice corrosion could be seen, especially on the throat and inner part of the platform; see Figure 2d. Especially the platforms and the upper part of the throat were affected by a strong corrosion attack. In this area, the contact surface between the root and the seat of the blades was filled with voluminous aluminum corrosion products and the gap was enlarged to about 0.7 mm.

In addition, the fracture surface on the blade side of the broken blade root was subjected to fractographical examination. Ultrasonic cleaning did not remove the corrosion products adhering to the fracture initiation area. In the area covered with corrosion products, which extended to approximately 1/3 of the fracture surface (red area in Figure 3b), the visual characteristics of the fracture were no longer recognizable: the fracture surface was destroyed by corrosion. The area that was not covered by corrosion products showed the appearance of a brittle cleavage fracture; see Figure 3c.

3.6. Metallography

Longitudinal metallographic sections were prepared through the suspected fracture initiation zone on the root of the fractured blade; see Figure 4. For comparison purposes, longitudinal metallographic sections were also prepared through the root of intact blades. The surfaces of all examined blades were affected in the areas of their throat and inner part of the platform by pitting, crevice corrosion, and cracks of 1–2 mm in length originating from defects, such as pores. The fracture surface of the broken blade was heavily corroded in the fracture initiation zone: small pits and crevices showed a localized corrosion attack; see Figure 4b. The secondary cracks extended preferentially along the interdendritic Al–Si eutectic; see Figure 4a,c. The characteristics of these cracks (path, branching) indicate that they were formed under the combined influence of corrosion and tensile stress on the part, i.e., it is stress corrosion cracking. The base material had a characteristic cast aluminum alloy structure consisting of primary aluminum dendrites, the Al–Si eutectic, and the intermetallic phase Mg₂Si; see Figure 4c. Compared to the refined structure of the material of the other examined blades, the structure of the broken blade was coarser. This was an indication of an unrefined or insufficiently refined structure. Spherical porosities of about 150 to 300 μm in size were homogeneously distributed in the part; see Figure 2d and Figure 4a. This relatively high porosity is typical for sand casting.

3.7. Clearance Measurement

The relevant dimensions of the assembly were measured to evaluate the risk of a mechanical contact between the end of the rotor blades and the inner surface of the housing. These measurements showed that the stiffness and circularity of the housing, the eccentricity of the rotor, and the variation in the length of the blades taken individually appear to be non-critical for the operation. However, the tip clearance (distance between the rotor blades and the housing) appeared to be small (in the range of just 0.8–5 mm) in comparison to the overall rotor dimensions (tip clearance ~1–5‰ of rotor diameter) and the structural compliance of the design.

3.8. Intermediate Conclusions

At this point of the investigation, it was clear that the purely mechanical static fracture, which in both cases could be clearly deduced from the fractographical examinations, remained inexplicable. While in the 2012 case, it was initially suspected that strong vibrations could have led to high loads, there were no signs of such vibrations in the 2015 case. For such a high load, the normal operating loads assumed when designing the rotor were far from sufficient. At least in the case of 2015, an unusual loading mechanism must have occurred that was effective enough to produce loads high enough to cause static quasi-brittle fracture.

4. Fracture Mechanics Assessment

The motivation to perform a fracture mechanics assessment was the need for an explanation of the fractographical finding—that the fracture structure showed a cleavage fracture in the uncorroded area of the fracture surface. The question to be answered was, Which loads could generate a stress intensity beyond fracture toughness? The fracture mechanics model was developed analytically based on standard solutions from the literature.

Concerning the load and thus the stresses in the component, qualitative considerations and order-of-magnitude estimates are made. This approach makes it possible to formulate a hypothesis on the failure of the component without having to quantify the stresses exactly. Therefore, three tasks had to be performed:

• Estimation of fracture toughness of the material;
• Determination of the normalized stress intensity of the critical crack front size;
• Determination of the stress level at the fractured cross section.

The stress intensity calculated for the stress level determined in step 3 could then be compared to the fracture toughness found in step 1.

4.1. Estimation of Fracture Toughness

The results of the tensile and notched bar impact tests were used to assess the material load-bearing capacity in the presence of a crack. For the estimation of the fracture toughness, a method was used that was developed for ductile structural steels but that nevertheless allows an estimation of the fracture toughness of the cast aluminum alloy. The fracture energy determined in a Charpy test at room temperature can be transformed into a fracture toughness value according to a correlation given by Schindler [10].

$$J_{0.2}=q\times KV,$$

(1)

where

$$q=1.7 \left[\frac{\frac{N}{\mathrm{mm}}}{J}\right],$$

(2)

and with

$$K_{Ic}=\sqrt{J_{0.2}\times \left(\frac{E}{1-{\upsilon }^2}\right)} \mathrm{MPa}\sqrt{\mathrm{m}},$$

(3)

using E = 76,000 MPa and υ = 0.33,

$$K_{Ic}\cong 12\mathrm{MPa}\sqrt{\mathrm{m}}.$$

(4)

This is an estimation only, because the fracture energy of KV = 1 Joule is low and the correlation of J_0.2 versus KV has been determined for values above 25 J and not for aluminum casting. However, it is close to the value of 12.67 MPa√m reported by Ranganatha for AlSi12 aluminum casting [11]. Estimating the uncertainty as ±25%, the fracture toughness of the material is in the following range:

$$9\mathrm{MPa}\sqrt{\mathrm{m}}<K_{Ic}<15\mathrm{MPa}\sqrt{\mathrm{m}}$$

(5)

4.2. Determination of Stress Intensity

The line in Figure 3b between the corroded and the non-corroded part of the fracture surface is the crack front when the final fracture occurred. In fracture mechanics, this is the critical crack size when the stress intensity at this crack front exceeds the fracture toughness. The stress intensity is a function of the stress level due to mechanical loading and a function of the geometry, e.g., the critical crack size.

$$K\left(\sigma ,b\right)=F_I\left(geometry\right)\times \sigma \times \sqrt{\pi b}$$

(6)

In our case, two solutions from the standard work by Murakami were used, which can be regarded as idealized limiting cases for the crack configuration found here. These are the solutions for “semi-elliptical surface crack in a long shaft under tension” and for “circumferential crack emanating from a circumferential notch of a cylindrical bar under tension” ([4] p. 654 & 646, 647). The lower-bound solution for the crack configuration is shown in Figure 5a and the upper-bound solution in Figure 5b.

For the lower-bound solution, the graphically estimated geometric parameters are 2a = 26 mm, b = 13 mm, and diameter of the cross section D₁ = 44 mm. This results in an approximate factor F_I value of 0.80, and the stress intensity per unit stress from a tensile load in the cross section is:

$$\frac{K}{\sigma }=F_I\sqrt{\pi b}=0.16\sqrt{\mathrm{m}}$$

(7)

For the upper-bound solution, the graphically estimated geometric parameters are c = 8.6 mm, D₁ = 44 mm, D₂ = 58 mm, t = 7 mm, and ρ = 10 mm. This results in an approximate factor F_I value of 2.325, and the stress intensity per unit stress from a tensile load in the cross section is:

$$\frac{K}{\sigma }=F_I\sqrt{\pi c}=0.382\sqrt{\mathrm{m}}$$

(8)

4.3. Determination of Stress Level in the Critical Cross Section under Service Loads

The load on the fractured cross section has two components: The first is the centripetal force due to the rotation of the rotor, which results in a tension stress, and the second part is a bending moment due to the aerodynamic pressure difference on the blade. With the mass of 1.9 kg rotating at 1500 rpm (f = 25 Hz), and the center of gravity at a distance of 0.3 m, we get the centripetal force as

$$F=m\times {\omega }^2\times r=m\times {\left(2\pi f\right)}^2\times r=14.1\mathrm{kN}.$$

(9)

With the cross sectional area A

$$A=\pi {\left(d/2\right)}^2=1770 {\mathrm{mm}}^2$$

(10)

we get a nominal tension stress σ_t of:

$${\sigma }_t=F/A=8 \mathrm{MPa}.$$

(11)

The bending moment due to the aerodynamic loads, which were determined by the designer of the jet fan, are as follows: An average pressure difference of about 1700 Pa acts on the area of the blade of A_b = 0.6 m², and with a bending arm a of 200 mm, the bending moment at the critical cross section is:

$$M_b=\Delta p\times A_b\times a=20,000 \mathrm{Nmm}$$

(12)

with W_b being the section modulus

$$W_b=\frac{\pi d^3}{32}=10,520{ \mathrm{mm}}^3,$$

(13)

we get a bending stress σ_b of

$${\sigma }_b=\frac{M_b}{W_b}=2 \mathrm{MPa}.$$

(14)

The resulting stress in the fractured section is therefore

$$\sigma ={\sigma }_t+{\sigma }_b=10 \mathrm{MPa}.$$

(15)

The local stress is higher because we have a notch, and using a stress concentration factor according to the solution “round tension bar with a U groove” in [12] with D = D₀ = 58 mm, d = D₁ = 44 mm, and ρ = r = 10 mm, we get a stress concentration factor K_t = 1.80, and the maximum local stress is:

$${\sigma }_{loc}=K_t\times \sigma =16.2 \mathrm{MPa}$$

(16)

This local stress is low: it is less than 15% of the static yield strength (R_p0.2 = 115 MPa). Based on an estimated fatigue strength endurance limit of 65 MPa (see Section 6.2), such a stress level should not result in a fatigue failure either.

4.4. Determination of Stress Intensity under Service Loads

Based on the stress intensity solution found in Section 4.2 we can now calculate the stress intensity under service loads as

$$K=\left(\frac{K}{\sigma }\right)\times \sigma =0.382\times 10=3.8 \mathrm{MPa}\sqrt{\mathrm{m}}$$

(17)

This stress intensity is far below the lower bound of the fracture toughness of 9 MPa√m of the material (see Section 4.1), and therefore a brittle fracture of the crack configuration found in the critical section cannot be explained with the service loads according to Section 4.3.

4.5. Intermediate Conclusions

Even by applying fracture mechanics, the failure cannot be explained. Two main observations found in the fractographical analysis remain unanswered: Why could stress corrosion cracking occur, and why could the root of the blade brake in a brittle manor? We can define three different factors that have to be present simultaneously to introduce stress corrosion cracking:

• A material sensitive to stress corrosion cracking by hydrogen embrittlement;
• Corrosive environment;
• High tensile stresses in the component acting for a long time.

The high silicon content (12%) and the copper content (1%) are significant with regard to the susceptibility to corrosion of this casting alloy. Aluminum casting alloys—with the exception of the aluminum–zinc–magnesium alloys—are generally considered to be not sensitive to stress corrosion cracking. Low susceptibility is shown by the copper-containing variants [13]. Aluminum–silicon alloys are generally not considered to be particularly susceptible to stress corrosion cracking [14]. However, laboratory tests show “environmental-induced cracking”, a type of stress corrosion cracking, in alloys with high proportions of silicon, magnesium, and copper (6061 alloy). Investigations were carried out under constant strain with the 6061 alloy in the heat treatment condition T4 (naturally aged) in a salt water–soda solution [7]. The mechanism behind the phenomenon of stress corrosion cracking in aluminum alloys is well known to be hydrogen embrittlement [15,16,17]. The presence of local corrosion pits, as found in the fractographical examination (see Figure 4b), provoke a local pH reduction, which then changes the local environment such that corrosion of the metal takes place and hydrogen is produced. The hydrogen itself then can diffuse in the exposed metal and make it brittle. From this, it can be deduced that in our case, the first two necessary conditions for stress corrosion cracking by hydrogen embrittlement are present.

However, the third factor is still open: High tensile stresses in the component are not obviously present, because the strength verification resulted in a local stress in the critical cross section of only 16.5 MPa. This stress level is low compared to the material strength. Moreover, this stress level is only present when the fan is operated. According to the operator’s report, however, the fans were only used in exceptional cases (total operating time of about 150–200 h in 9 years). During standstill, the stresses in the blade root are negligible. Searching for a scenario with high tensile stresses for a long time brought up the following hypothesis.

5. Hypothesis for High Tensile Stresses in the Root of the Blades

The findings have shown that the gap between the blade platform and the upper surface of the mounting block is completely filled with oxidation products (Figure 1 and Figure 2). These oxidation products have, compared to the corresponding mass of metal, a larger volume. Thus, one can imagine the following process: At the beginning, i.e., when the rotor is assembled, a fine gap remains between the blade platform and the mounting blocks. Being then exposed to the corrosive environment in the tunnel, the corrosive medium begins to accumulate in the gap, and over time, the aluminum alloy, which has no surface protection in the gap, oxidizes. The oxidation in the gap leaves oxidation products completely filling the gap and gradually creating a wedge effect on the surrounding parts. The forced widening of the gap causes a displacement of the platform of the blade, resulting in a tensile elongation in the throat of the blade (Figure 6).

From a mechanical point of view, this results in the following situation: The root of the blade is in contact with the mounting blocks in the area of the smallest cross section. In the area of the platform, a pressure acts on the root of the blade on the side of the gap. The throat of the root is thereby subjected to tensile stress like a short tension rod (the flow of force path from the gap over the upper part of the root back to the support points in the smallest cross section). Forced tensile stresses arise in the throat. The order of magnitude can be estimated with the following consideration of the geometric conditions: The actual gap size is about 0.6 mm. Assuming an initial gap of 0.45 mm, the widening of the gap and thus the lengthening of the shaft are ΔL = 0.15 mm. In relation to its height L of approx. 17 mm, this corresponds to an imposed elongation of 0.9%. As the elongation at break determined in the tensile test is about 0.8–1.45%, the above-estimated elongation is in the range of the maximum elongation of the alloy and can therefore cause failure. Correspondingly, the tensile stresses reach values up to or beyond the yield strength. Assuming a flow stress of 162 MPa (average of R_p0.2 and R_m) for the stress in the throat, we can conclude the following: the widening of the gap due to the increase in volume as a result of the formation of oxidation products has the potential to cause tensile stresses reaching the load-bearing capacity of the material. With this mechanism, we have all three factors necessary to introduce stress corrosion cracking.

Furthermore, a brittle failure can be explained: With a stress in the throat equal to the flow stress, we get a stress intensity of at least

$$K=\left(\frac{K}{\sigma }\right)\times \sigma =0.16\times 162=26 \mathrm{MPa}\sqrt{\mathrm{m}}$$

(18)

This stress intensity is beyond the upper limit of the fracture toughness of the material, whereby the stress due to the wedging effect described before is able to bring the crack front to instability, i.e., to trigger a brittle fracture.

6. Assessment of the Corrosion, Fatigue, and Corrosion Fatigue Properties of Aluminum Alloys

Three principal options were identified as potential corrective actions: design improvements, change of the material, or improvement of the surface protection against corrosion.

As improvements in design are in the competence of the jet fan supplier, this possibility is not discussed.

For the second option two aluminum casting materials were characterized: AlSi12CuMgNi according to EN AC-48000 and AlSi9Mg according to EN AC-43300. The second alloy was a candidate as a replacement material. This material has no copper and less silicon, and the question was whether this would result in better stress corrosion resistance under a typical corrosive environment in a tunnel.

The third option was better surface corrosion protection. Powder coating or wet coating can significantly inhibit a corrosion attack. While the rotor of the first failure case had no surface protection, the jet fan of the second failure event had a surface coating, but it was not properly applied: part of the rotor blades and the mounting blocks were not covered, which allowed corrosion to occur in these areas. It had to be investigated whether surface protection before assembly of the parts of the rotor is possible and is sufficient for long-time protection against corrosion.

6.1. Evaluation of the Corrosion Resistance

To evaluate corrosion resistance, corrosion chemical parameters were defined first. Four different aluminum alloys were considered: AlSi12CuMgNi (EN AC-48000-T5), AlSi1MgMn (EN AW-6082-T4), a wrought alloy AlSi9Mg (EN AC-43300-T5), and AlSi10MnMg (EN AC-43500-T5) both casting alloys. The sample surfaces were degreased with ethanol and sanded with 600 grit sandpaper before testing. The potential-dynamic current density–potential curves were measured with a potentiostat-type Jaissle PGU 10V-1A-IMP-S (Ingenieurbüro Peter Schrems, Münster, Germany) with a 3-electrode cell with a volume of about 700 mL, a Pt counterelectrode, a scan rate of 1 mVs⁻¹, and a contact cell of 1 cm² area (defined using an EPDM O-ring) against a mercury/calomel/KCl saturated (SCE) reference electrode in a 0.5 wt% sodium chloride solution at 23 °C. The surface of the sample was contacted in such a way that a dense cell could sit on the surface. The data were recorded from −100 mVSCE (relative to the rest potential, a conversion of the potential values in relation to the standard hydrogen electrode can be performed by adding an offset of 244 mV) up to a corrosion current density of >1 A/m²; see Figure 7. The rest potential, free corrosion current density at the rest potential, and pitting corrosion potential were determined; see

Table 2. Corrosion chemical characterization of the aluminum alloys investigated—typical values of aluminum (99%) included for comparison.

Material	Rest Potential (mVSCE) *	Free Corrosion Current Density at Rest Potential (µA/cm2)	Corrosion Velocity ** (mm/a)	Pitting Corrosion Potential (mVSCE) *
AlSi1MgMn (6082-T4)	−860	3	0.03	−720
AlSi9Mg (EN AC-43300-T5)	−760	2	0.02	−760
AlSi10MnMg (EN AC-43500-T5)	−730	2.5	0.03	−730
AlSi12CuMgNi (EN AC-48000-T5)	−680	9.5	0.1	−720
Al (99 %)	−780	0.6	0.006	−650

* Measured against a mercury/calomel/KCl saturated (SCE) reference electrode. ** Calculated based on the free corrosion current with 0.0109 mm/a/μAcm².

.

Based on the free corrosion current, the corrosion velocity in millimeteres per year was also calculated by assuming a corrosion velocity of 0.0109 mm per year per unit current density. Pure aluminum is included in the table for comparison. In the case of the wrought aluminum alloy (AlSi1MgMn), a passive material behavior (relatively flat rise of the curve) was recognized in the anodic branch of the current density (see Figure 7; branch on the right). In the case of the aluminum casting alloys, no or only insignificant passive behavior could be detected, i.e., these alloys showed active corrosion behavior in chloride-containing media. The difference in the passive behavior of the wrought aluminum alloy and the cast aluminum alloys is probably due to the different manufacturing process: Cast alloys have pores, blowholes, and a significantly more inhomogeneous microstructure than wrought alloys, so there are always places on the surface that can be activated. The cast aluminum alloy AlSi12CuMgNi (actual blade material of the failed jet fans) showed the most unfavorable material behavior among the samples examined, since the highest corrosion current could be measured at the rest potential. A surface corrosion rate of 0.1 mm/a could be calculated for this material. The aluminum casting alloys AlSi9Mg and AlSi10MnMg had a corrosion rate that was smaller by a factor of 10 compared to AlSi12CuMgNi. These values can only be taken for a qualitative comparison, since in practice, local corrosion phenomena, such as pitting corrosion, and in the case of mechanical tensile stress, stress corrosion cracking must also always be expected with aluminum materials. The aluminum casting alloys AlSi9Mg and AlSi10MnMg have therefore better corrosion resistance in a tunnel environment than the alloy from the fractured blade; however, passivation behavior cannot be expected. The casting alloy AlSi9Mg was further evaluated as an alternative to the alloy AlSi12CuMgNi.

6.2. Evaluation of Fatigue Properties

As fatigue strength estimated from static data is not reliable, fatigue tests were performed. Fatigue strength data were determined for the two casting alloys, AlSi9Mg and AlSi12CuMgNi, in rotation bending (stress ratio R = −1).

For the fatigue and corrosion fatigue specimens, material was taken from new blades (electro-eroded and machined) and prepared with a surface finish of standard machining with a roughness class of approx. N5 to N6 and a center roughness value of R_a = 0.4–0.8 μm. The test procedure was according to DIN 50100 [18] and DIN 50113 [19] and performed on a Schenk Rapid PUN 052Z (Carl Schenk AG, Darmstadt, Germany). The rotating bending tests were carried out with constant load amplitudes and a constant test frequency of f = 100 Hz. The tests were carried out until specimen breakage or were stopped after reaching 10⁷ load cycles. From these tests, an endurance limit for a horizon of 5 × 10⁶ cycles (run-outs) was estimated for both aluminum alloys: AlSi12CuMgNi and AlSi9Mg had almost the same endurance limit, namely 65 MPa and 67 MPa, respectively.

6.3. Evaluation of Corrosion Fatigue Properties

To determine the corrosion fatigue characteristics of the alloys, a specific test was designed and a corresponding test bench was constructed. The tests consisted of bending the round specimens with a frequency of 0.2 Hz. The corrosive environment was generated, on the one hand, by periodic, weekly spraying with a 0.5 wt% chloride-sodium solution. On the other hand, each specimen was in a closed chamber with a deposit of approx. 20–30 mL of this solution so that a humid, corrosive climate was established in the test chamber; see Figure 8a. The tests revealed a reduction in service life by a factor of 2.0 and 1.6 for AlSi12CuMgNi and AlSi9Mg, respectively, in a corrosive environment. The corrosive influence therefore does not differ significantly between the two materials. In addition, based on the estimated fatigue strengths of both materials without a corrosive influence, a comparable, fatigue-resistant bending stress amplitude can be assumed. Since the operating stresses of the fan blades, apart from start-up and shut-down processes, probably have rather small but higher-frequency stress amplitudes (vibrations with high numbers of cycles), the two materials are classified as almost equivalent in terms of fatigue corrosion resistance.

In contrast, the two materials differed significantly in the short-term strength range, i.e., at bending stress amplitudes around 100 MPa; see Figure 9. In this regime, the material AlSi9Mg showed a significantly longer fatigue life.

7. Discussion

Concerning the load on the blade and thus the stresses in the blade root, the circumstances allowed only an analysis with uncertainties. A more precise quantification of the tensile stresses due to the wedging effect is not possible. The geometrical condition regarding the initial gap is unknown and cannot be reconstructed. The uncertainties accepted in this approach have no influence on the conclusion, and we do not want to give an exact quantitative answer but only a qualitative one with some order of magnitude of the variables involved. This nevertheless allows the following hypothesis for the failure of the component:

Since the installation of the jet fan in the tunnel, the corrosive medium infiltrated the gap between the throat of the blade root and the mounting blocks. As the individual parts were not properly protected by surface treatment, corrosion occurred. The resulting oxidation products filled, over time, the gap completely and increasingly stretched the blade at its root and therefore caused forced tensile stresses. Furthermore, the presence of a corrosive medium provided the necessary aggressive environment to trigger stress corrosion cracking, i.e., environmental-induced cracking in the transition radius from the throat to the platform of the blade root. The crack developed over time into a surface crack over about ¾ of the circumference of the shaft. A critical crack front configuration was reached at about 11 mm crack depth, and the forced tensile stresses triggered a fast brittle fracture.

As the cause of the failures comes down to materials and design, in combination with a corrosive environment, the problem is potentially general to all jet fan rotors of this type installed in Switzerland (and other countries in which dew-salts are used). The variability in the geometry of the mounting blocks and the blade, caused by the manufacturing process (casting), suggests that the forced tensile stresses can vary greatly from blade to blade. Therefore, no general rule can be derived when the critical operational life is reached. Furthermore, the presence as well as the size of eventual cracks cannot be identified visually without disassembling the fan and the rotor. The unknown risk of further failures and with that the risk of severe traffic accidents led to the replacement of all the rotors of this type.

In our case, the mechanical service loads and the chemical environment of the component were such that the degradation could be driven by corrosion mechanisms or fatigue or even a superposition of both. The case presented here is significant in the sense that it was not an obvious design fault that induced degradation processes ending in a failure of the component. It was more a combination of circumstances under which a mechanism could take place, which was not considered in the design rules: a static stress corrosion cracking phenomena induced by wedging of corrosion products in a narrow interface gap between aluminum parts.

8. Conclusions

Fractography showed a corrosion attack at the interface of the blade root and the clamps. Crevice as well as cracks as deep as 13 mm were found. The fracture surface also showed a final corrosion-free brittle static fracture surface. The cause of the final failure must have been a purely mechanical static overload. The operating loads could by no means generate stresses high enough to initiate fracture. A wedging mechanism induced by corrosion products was found to be effective enough to produce loads high enough to cause static brittle fracture. The damage scenario resulting in the fracture of the rotor blade is as follows:

(1)

Formation of a crack

The unprotected interface between the blade root and the clamps was exposed to the aggressive corrosive environment: the AlSi12CuMgNi casting alloy shows active corrosion behavior. The corrosion products formed between the blade root and the clamps led to a wedging effect building up tensile stresses in the blades root. Therefore, crack formation occurred (stress corrosion cracking).

(2)

Propagation of the crack

The continuous presence of a corrosive medium and therefore the oxidation product formation in the crack increased the wedging effect, and therefore the crack propagated further (corrosion fatigue process).

(3)

Final fracture

The sudden failure of the blade root was a static, brittle, cleavage fracture occurring under the influence of the tensile stresses based on exceeding the fracture toughness of the material.

(4)

Correction actions

As complementary tests showed that other aluminum casting alloys do not present significant advantages regarding corrosion fatigue, the most interesting option appears to be proper corrosion protection of each individual aluminum alloy piece. Therefore, such components must be fully protected against corrosion with a two-layer powder coating with a total nominal coating thickness of 170 μm or a three-layer wet coating with a total nominal thickness of 160 μm, as specified in [2]. As no design change is implied, compliance with other regulations, such as the fire rating, is not impacted by this corrective action and existing rotors can easily be replaced.

As part of maintenance, regular inspection has to be planned, which can effectively detect degradation of the corrosion protection and initiate repair, if necessary.