Corrosion of the Carola Bridge

Second and final part: partial collapse, stress corrosion cracking, chloride stress

The collapse of the Carola Bridge in Dresden on September 11, 2024 is certainly the most significant case of corrosion-related material failure in recent times and also serves as a prominent example of the country's ailing infrastructure. The partial collapse of the prestressed concrete bridge occurred without prior notice. Originally, the Carola Bridge was considered an icon of engineering of its time. It was a very aesthetic and slender structure. Its design and construction would still be a challenge today. The existing documents testify to a high quality of design and very careful construction. Following the first part of the article in Galvanotechnik 5/2025, the second part now deals with the collapse, the attempt to reconstruct the collapse process and the search for the cause of the collapse.

Partial collapse in September 2024 and research into the cause

On the night of September 10-11, 2024, the bridge c of the Carola Bridge collapsed. Fortunately, there were no streetcars, pedestrians or cyclists on the structure at the time of the partial collapse. The traffic load was therefore zero. The findings, which were available by the end of 2024 and which were largely presented at the public meeting of the City of Dresden's Building Committee, are presented below [23].

Reconstruction of the collapse

Investigations into the cause of the partial collapse of the Carola Bridge are now well advanced. Various hypotheses have been tested and comprehensive diagnostic and computational analyses have been carried out. The numerous reports from the public and experts have also been meticulously investigated. The interim results indicate that hydrogen-induced stress corrosion cracking was the main cause of the failure of the tendons and subsequently of the entire structure.

It can be assumed that a not inconsiderable proportion of the prestressing wires in tension c had already been damaged for some time, and in some cases broken. This is suggested by the findings presented in the following sections. The resulting loss of stiffness of superstructure train c led to the cross-connection to train b at the level of joint II being subjected to increasing stress. It can be assumed that it was precisely this cross-connection and the resulting possible load transfer to neighboring bridge sections that may have prevented a collapse for a long time.

Changing traffic loads and temperatures were primarily responsible for the constantly changing state of stress. A particular superposition of climatically induced temperature stress and traffic load probably led to the decisive tension wire breaks, resulting in further load redistribution and thus to the overloading of the cross girder connection in the morning hours of September 11, 2024. At the moment of failure of the articulated cross connection, the support cross-section was no longer able to absorb the resulting stress.

In the two weeks before the collapse, it was unusually hot in Dresden. In the night from September 10 to 11, the air suddenly cooled down considerably. The top of the bridge shortened. At the same time, the Elbe water continued to heat the bridge from below. This led to high vertical temperature differences and to an upward movement of the cantilevers of the girders. This movement can be traced, for example, in the data from the joint gap measurements. The resulting residual stresses at tendon level were somewhat greater for tension c than for tension a, as tension c did not have a buffering asphalt layer that could have dampened the temperature effect.

The upward movement led to a relief of the transverse connection and thus a reactivation of prestressing steel above support D. As a result of the traffic load from the passing streetcars, which counteracted this movement, further prestressing wires broke. The cantilever of train c increasingly relied on the cross-connection to bridge train b. Ultimately, the bracket was overloaded and tore off, which immediately led to the failure of superstructure c in axis D. The picture on the left shows an aerial view of the structure shortly after the collapse.

The structural analyses of the Carola Bridge confirmed that the very slender structure was designed, constructed and built with a very high cross-sectional utilization. These special features prevented the formation of pronounced cracks that could have heralded the failure. Either the cracks are not recognizable due to the extremely heavily reinforced tension chord at axis D with a barely noticeable increase in elongation when individual steels fail, or the crack widths remain so small according to plan that the announcement of a critical condition for the stability would in fact not take place at all or only much too late. A reliable prediction, e.g. according to the crack-before-fracture criterion [24] [25], would therefore not have been possible with the permissible and usual methods.

Stress corrosion cracking of the Carola Bridge

Stress corrosion cracking (SCC) is considered to be the main initial cause of the collapse of the bridge c of the Carola Bridge. Basically, hydrogen-induced stress corrosion cracking can occur on high-strength prestressing steels. However, other parameters must also be present. The decisive factor here is usually corrosion processes on the prestressing steels. A mandatory prerequisite for this is that the so-called electrolyte (e.g. a raindrop, condensate or other moisture source) has a pH value < 5. If this is the case, the cathodic partial reaction of the corrosion process takes place with the formation of hydrogen, as a result of which cracking can occur if voltage occurs at the same time. Improper handling of the steels during transportation or on the construction site as well as exposure to pollutants such as chlorides can also be critical in this context. The subsequent crack growth due to the mutual stimulation of stress and corrosion, possibly accompanied by a reduction in cross-section despite an alkaline environment due to the grouting mortar, can last for a few days or even many years until the steel finally fails due to residual stress fracture of the remaining cross-sectional area of the affected wires.

Hydrogen-induced stress corrosion cracking is a relatively rare type of corrosion. It can generally occur on high-strength steel materials (tensile strength _Rm > 800 MPa) [26]. It is linked to the presence of specific prerequisites. These include a material that is particularly susceptible to this type of corrosion due to its composition and manufacture, which applies to oil-tempered prestressing steels, the presence of an electrolyte with a pH value < 5 and mechanical tensile stress [27]. All of these conditions were present in the Carola Bridge.

It is very likely that the mechanism of the SpRK was set in motion despite the conscientious handling of the material during construction and that a chain of several circumstances ultimately led to this overall result. In addition to the non-existent ability to redistribute forces in the statically determined individual supporting structures, this also includes the special conditions of manufacture. The continuous web tendons in the cantilever beams were prestressed and grouted immediately afterwards. The subsequent tendons, however, which were mainly located in the roadway slab and were intended to absorb the later loads of the suspension girders, were initially partially prestressed and remained ungrouted in the duct during the construction of the current suspension girder. It is very likely that during this time a condition arose, e.g. due to condensation in the cladding tube, which allowed condensation to form at the high points. The release of heat from the hardening concrete, especially from the solid webs, as evidenced by the built-in cooling pipes, certainly had a beneficial effect with a warm and humid climate in the box girder. It is known that even small amounts of water (electrolyte) in the form of droplets on the prestressing steel surface are sufficient to initiate stress corrosion cracking. The conditions for the occurrence of H-induced stress corrosion cracking were present in this structural condition over a longer period of time. The thesis on the exposure conditions of the long exposed prestressing wires in the unpressed cladding tube (especially in the deck slab) is supported by the damage rates of almost 80 % and more observed.

Earlier investigations by the Federal Institute for Materials Testing (BAM) had shown that even a brief exposure of the Hennigsdorf prestressing steel to moisture was sufficient to initiate the SpRK mechanism and the first cracks. As a result, the requirements according to [21] to grout prestressing wires in ducts within ten days without applying special protective measures cannot prevent crack initiation in the event of condensation. The process comes to a standstill after grouting, as the pH value is sufficiently high due to the alkaline grouting mortar. Corrosion processes to form a protective oxide layer on the prestressing wires now take place with the reduction of oxygen; no hydrogen development takes place. After the formation of this oxide layer and with the cladding tube intact, no corrosion occurs on the prestressing wires. However, incipient cracks represent the initial damage, which can extend through fatigue stresses until the prestressing steel breaks. This scenario is plausible for the Carola Bridge. The stress caused by streetcars is an important indicator here. While the surface loads of a streetcar are relatively low, the relatively high axle loads had a direct and immediate fatigue effect on the bridge span c. Due to the structural design, this had a particularly critical effect on the heavily loaded cross-section in the pier axis D with its highly loaded compression zone, where a progressive prestressing steel failure ultimately initiated the final failure.

In the hours following the collapse, as much as possible was documented in the area of the fracture. Photos of the fracture site proved to be particularly valuable. Figure 1 shows an excerpt from the construction documents in the upper section. The condition of the broken bundle tendons was marked in color. A very diverse picture was found. The tendons in the deck slab in particular had been damaged for a long time. This can be seen from the extremely dark fracture surfaces of many of the tendons. These fractures must have occurred during construction or shortly afterwards, but in any case some time ago (bottom left in Fig. 1). During microscopic examinations of the fracture surfaces of broken wires of the Carola Bridge at the Federal Institute for Materials Research and Testing (BAM) in Berlin, mortar adhesions were found in isolated cases, which are a clear indication that the fractures in question must have occurred before or during grouting. The prestressing steel surface intact at the time of the partial collapse in cross-section D was determined to be 0 to 20 % in the roadway slab area for more than 2/3 of all tendons. The BSGs in the webs were essentially less damaged, with the condition downstream being better than upstream. It should be noted that these do not show a ductile fracture pattern and therefore almost all broken prestressing wires in axis D exhibit embrittlement of the material. Wires with cracks exhibited the cracks typical of prestressing wires. Freshly broken surfaces were shiny metallic without traces of corrosion (Fig. 1).

As can also be seen from the images of the fractures, the installed slack reinforcing steel was probably pulled out of the concrete over long distances without major resistance, which clearly indicates a bond failure. The reasons for this are not yet completely clear. Some of these could go back to the time when the bridge was built, if the reinforcement was walked on during concreting or already during concrete hardening or if the concreting could hardly be carried out without cavities from the outset due to the extremely narrow reinforcement layers caused by the large number of prestressing and reinforcing steel reinforcement. The cracks detected parallel to the tendon layers, which will be discussed later, could also have permanently damaged the bond strength. This is all the more regrettable because the verifications of the announcement behavior contained in the current guidelines rely heavily on the load-bearing effect of the reinforcing steel not affected by stress corrosion cracking.

Fig. 1: Results of the photographic documentation of the tendon condition in the fracture cross-section axis D tension c (as of mid-December 2024) BSG: Bonded tendons - (Graphics and photos: MKP GmbH, [23], mod.)

Structural monitoring until 2024 Report on measurements at the beginning of the 1990s [12]

A prerequisite for the stability of the bridge was sufficient pretensioning of the coupler bolts with which the cast steel joints were connected to the web tendons in the longitudinal direction of the bridge. Of the total of 504 bolts in the nine joints, 121 were therefore designed as controllable measuring bolts.

The use of mechanical precision dial gauges was expected to provide the highest reliability in force measurement. The existing bolt force was determined from the difference in elongation between the zero measurement and the measurement under load. The bolt forces were determined during production and then again in 1974, 1979, 1982 and again in the early 1990s.

As early as 1983, a clear deflection was observed at joint II, which had increased further in the meantime. The measurement campaign at the beginning of the 1990s was intended to investigate and verify the causes and the effects on the use of the structure. In the box girder of train c, inclinations, displacements, vibrations and temperatures were recorded with various measuring devices at different positions. Due to seasonal fluctuations, some of the measurements were taken over a period of more than a year. The most important results are presented briefly.

The absolute values of the coupling bolt forces fluctuated greatly and decreased continuously since they were preloaded.

However, the decrease slowed down over the years. The deformations responsible for the decrease in forces were attributed to creep and shrinkage of the concrete in the connection area. Damage to the prestressing steels was not observed at the time. This theory was followed for a long time with appropriate evidence and no reason was found to raise the question of sufficient stability.

Depending on the load position, the passage of a tramway caused a deflection of joint II of up to 12 mm downwards and 3 mm upwards. This was not relevant for the global load-bearing behavior. The deformations due to climatic temperature differences were considerably greater and reached a span of up to 65 mm within 1.5 days. Compared to the zero position, the deflections amounted to between +20 mm (upward movement) and -80 mm (downward movement). On the basis of the available measured values and theoretical considerations, it was estimated that Joint II had moved downwards by an average of 30 cm by the beginning of 1993. This would correspond to approx. 80 % of the total average deformation to be expected by the end of the service life after 80 years. Critical consequences for the overall structure were not derived.

Announcement behavior

The topic of stress corrosion cracking is not new [28-30]. The cases of damage to the Berlin Congress Hall in 1980 [31] and the collapse of a truss in a production hall in Mannheim in 1989 [30] are known from the 1980s. As a result, various adjustments were introduced both in the production of prestressing steel and in the design rules to ensure sufficient robustness of prestressed concrete structures. Furthermore, great importance was attached to the existence of an announcement behavior of such structures, e.g. [24], in order to prevent a collapse-like failure in the event of unnoticed prestressing steel failure. Due to cases of damage, the focus was initially on the quenched and tempered prestressing steels Neptun N40 (Felten & Guilleaume Carlswerke AG) and Sigma (Hütten- und Bergwerke Rheinhausen AG). Since the mid/end of the 1990s at the latest, it has also been known that the Hennigsdorfer prestressing steel used in the Carola Bridge, which is also quenched and tempered, is sensitive to this type of corrosion [32]. It was deliberately included in the list of endangered materials in 2011 in accordance with the instructions on stress corrosion cracking [25]. The fundamental problem for the Carola Bridge was known. Regular structural inspections were carried out in accordance with standards, initially without any abnormalities. In 1996, a first recalculation was carried out according to [24], in which the calculated announcement behavior could be verified.

Transverse cracks in the roadway slab were documented for the first time in 2000. Two main causes are conceivable: excessive creep deformations or failure of individual prestressing wires in the bundle tendons (e.g. as a result of SpRK). As the creep deformations were known due to the observed deflections of the structure, these were considered to be the main cause of the cracks. This was plausible. The possibility of unnoticed prestressing steel fractures was not considered. The detected cracks had very small crack widths and were not considered to be significant. Since the announcement behavior was calculated, larger cracks were expected in the event of an actual failure pre-announcement.

The barely perceptible increase in strain due to the extremely high degree of reinforcement and the limited rotational capacity, particularly of the support cross-section in axis D due to the high load, were not considered in detail. Similarly, the relieving load-bearing effect in the transverse direction due to the articulated transverse connection at joint II was underestimated or not taken into account. The alternating support of the superstructures inevitably led to a small change in crack width in the event of critical damage to a tension.

The calculation of the fatigue strength of the prestressing steel did not reveal any deficits. Other measures carried out included measuring the crack movements in order to substantiate the observations on fatigue strength. Joint gap measurements were carried out on joint II from 2004. From today's perspective, however, these were unsuitable for the assessment of pre-announcement behavior.

Chloride stress

It should be mentioned that parts of the three superstructures, particularly between axles D and E, also exhibited damage as a result of chloride-induced corrosion. A damaged drainage pipe was considered to be the main cause at the time. This damage was repaired by means of non-destructive electrochemical chloride removal (CITec Concrete Improvement Technologies GmbH, Dresden) and the surfaces inside the box girder were additionally sealed to bridge cracks. Corroded prestressing wires and slack reinforcements were found during the structural investigation in the fall of 2024. However, these were not the cause of the failure.

Structural investigation since fall 2024

In recent months, a large number of material investigations have been carried out in situ on trains a to c and in the laboratory. In some cases, various non-destructive testing methods were used in order to compare their informative value. In addition to investigating the causes of the collapse, further objectives were pursued. The information about the material and the condition of the building should be used to

• clarify the cause of the damage and the course of the collapse,
• enable an assessment of trains a and b, e.g. with regard to their (temporary) continued use, and allow conclusions to be drawn about other structures with similar problems,
• realistic condition information can be obtained, which would be relevant, for example, for a dismantling analysis.

Concrete measures since then have included applying a strain gauge to a wire to measure the residual stress before it is cut in order to be able to draw conclusions about the existing stress from the recovery (Fig. 2). This resulted in a wide range of strains, from no strain to residual strains in the expected range.

Fig. 2: Sampling point in the floor slab of the superstructure train b, strain gauge for measuring the back strain on the left and cutting through the instrumented wire using a Dremel on the right. DMS: Strain gages - (Photos: Silke Scheerer)

Magnetic particle tests were also carried out on wires of the Carola Bridge at the BAM in Berlin. Figure 3 above shows one of these wires under UV light. An enormous number of cracks can be seen, which can be clearly attributed to stress corrosion cracking. At the bottom of the picture is a longitudinal section of a tension wire, which also shows the typical cracks very clearly. Of the 37 samples examined by mid-December 2024, 25 showed cracks. The majority of the wires examined from tension c were found to have such previous damage.

Further measures included drill core removal, opening of tendons and removal of grout and prestressing wires, remanence magnetism and tensile tests on prestressing steel.

Fig. 3: Cracking test using fluorescent magnetic powder (top) and longitudinal section (bottom) on two wires removed from the Carola Bridge - (Photos: BAM Berlin, [23])

Conclusion and outlook

The bridge was tested and monitored in accordance with the recognized rules of engineering and yet the failure occurred without warning. Consequently, it is necessary to critically scrutinize the rules for monitoring and testing and - if necessary - to adapt them. It is difficult, if not impossible, to make statements about the inside of the component. It is not possible to accurately and comprehensively assess the condition of the prestressing steel using the usual methods for structural inspections as long as no external changes such as cracks on the surfaces of the prestressed concrete components or atypical deformations of the components are visible. However, knowledge of the condition of the prestressing steel would be extremely important, as the partial collapse of the Carola Bridge and other prestressed concrete bridges currently under observation or being dismantled show.

Therefore, the recognizability of cracks as an indication of an imminent failure is of particular importance, regardless of the crack width. The instruction on stress corrosion cracking [25] logically defines that non-visible cross-sections, e.g. in the support area, if direct crack detection is not possible there due to coverings or elastic coatings, are considered to be undetected cross-sections in the sense of announcement behavior. Fiber optic monitoring of the support areas could provide a remedy here. Such distinctions were not yet included in the recommendations for the verification of announcement behavior [24], which were used for the Carola Bridge.

The computational pre-announcement behavior is used to verify whether the prestressed concrete component can develop sufficient load-bearing reserves to avert a collapse-like failure even in the event of unnoticed and gradual prestressing steel failure. The reinforcing steel reinforcement, which is not sensitive to SpRK, plays a decisive role here in order to be able to build up as sufficient a residual load-bearing capacity as possible with proper anchoring. Cracks running parallel to the reinforcement inserts can limit the bonding effect just as much as a too thin concrete cover, which is why the detection of parallel cracks and the measurement of the concrete cover should be given a more important role for the announcement behavior.

Longitudinal cracks became noticeable in the Carola Bridge, the cause of which has not yet been fully clarified. In addition to temperature constraints in the box cross-section, ring tensile forces due to back-anchoring of broken prestressing wires could also be responsible. Such cracks have also been reported in other structures with stress corrosion cracking and have been associated with back-anchoring, such as the Elsenbrücke bridge in Berlin and the Altstädter Bahnhof bridge in the city of Brandenburg an der Havel [34]. The instructions, however, focus on the formation of bending cracks and, to a lesser extent, longitudinal cracks. Measurement methods, such as acoustic emission monitoring for the detection of prestressing wire fractures, have been known for some time, but have largely remained a special solution in the context of object-related monitoring [35]. With the knowledge of the brittle failure risk of the prestressing reinforcement, it is not justifiable to put trains a and b (road bridge) back into operation. This decision was discussed and weighed up in many ways, as the Carola Bridge was an essential part of Dresden's road network. Among other things, a test load was considered. However, damage patterns similar to train c were found in both superstructures of trains a and b, indicating an advanced state of damage in these substructures as well. In addition, due to the feared extremely brittle failure mechanism with the hydrogen-induced stress corrosion cracking present here, test loading was considered too risky.

The complete dismantling of the Carola Bridge cannot be avoided. The planning and implementation of a new construction are currently being prepared by the city of Dresden. The aim is to rebuild the bridge as quickly and economically as possible. A design competition is also under discussion.

Literature
[23] Recording of the special meeting of the building committee on 11.12.204, available at https://www.dresden.de/de/rathaus/politik/stadtrat/stadtratssitzung-live3.php and presentation given there by Steffen Marx: Carolabrücke - Ursache und Hergang des Teilinsturzes. MKP GmbH, 2024.
[24] Federal Ministry of Transport (BMV, ed.): Recommendations for the inspection and assessment of bridge structures built with quenched and tempered prestressing steel St 145/160 Neptun N40 up to 1965. 07/1993.
[25] Federal Ministry of Transport, Building and Urban Development (BMVBS, ed.): Instructions for the inspection and assessment of older bridge structures built with quenched and tempered prestressing steel at risk of stress corrosion cracking (Handlungsanweisung Spannungsrisskorrosion). 06/2011.
[26] Kuron, D. (ed.): Hydrogen and Corrosion: Commemorative publication on the 60th birthday of Prof. Dr. Hubert Gräfen. Published in: Bonner Studien Reihe 3, Bonn: Verlag Irene Kuron, 1986.
[27] Nürnberger, U.: Korrosion und Korrosionsschutz im Bauwesen. Bauverlag, 1995.
[28] Analysis and evaluation of damage to prestressing steels. Research, Road Construction and Road Traffic Engineering 308 (1980), pp. 1-195.
[29] Mietz, J.; Nürnberger, U.; Beul, W.: Untersuchungen an Verkehrsbauten aus Spannbeton zur Abschätzung des Gefährdungspotentials infolge Spannungsrisskorrosion der Spannstähle. Final report on the BMV research project FE 15.209 R91P, 2 parts: BAM reference: Vh 1341 and FMPA - No. 34-10566, Berlin/Stuttgart, 1994.
[30] Nürnberger, U.; Beul, W.: Abschlussbericht zum BMV-Vorhaben FE 15.209: Untersuchungen an Verkehrsbauten aus Spannbeton zur Abschätzung des Gefährdungspotentials infolge Spannungsrisskorrosion der Spannstähle, Teil 2 - Untersuchungen der FMPA (FMPA-Nr. 34-10566), Stuttgart, 1994.
[31] Schlaich, J.; Kordina, K.; Engel, H.-J.: Teilinsturz der Kongresshalle Berlin - Schadensursachen, zusammenfassendes Gutachten. Beton- und Stahlbetonbau 1980 (75) 12, pp. 281-294.
[32] Mietz, J.; Fischer, J.; Isecke, B.: Prestressing steel damage to a bridge structure due to stress corrosion cracking. Beton- und Stahlbetonbau 93 (1998) 7, 195-200 - DOI: 10.1002/best.199800370
[34] Kaplan, F.; Steinbock, O.; Saloga, K.; Ebell, G.; Schmidt, S.: Monitoring of the bridge at Altstädter Bahnhof. Structural Engineering 99 (2022) 3, pp. 222-230 - DOI: 10.1002/bate.202200008.
[35] Käding, M.; Marx, S.; Schacht, G.: Acoustic emission monitoring for tension wire breakage detection. In: Bergmeister, K.; Fingerloos, F.; Wörner, J.-D. (Eds.): 2023 BetonKalender, Berlin: Ernst & Sohn, 2023, pp. 745-777 - DOI: 10.1002/9783433611180.ch15, 2022.