CHAPTER 15 STRENGTH EVALUATION OF THE VISOR AND THE RAMP ATTACHMENTS

The bow visor structure was built to scantling requirements specified in the Bureau Veritas Rules of 1977. Compliance with these has not been verified in detail in this investigation.
The locking devices should, according to the Bureau Veritas rules valid at the time, cause the bow door to be “firmly secured”. Structural reinforcements were specified in general wording to be required at attachment points for cleats, hinges and jacks.
Thus the Bureau Veritas rules did not specify minimum pressure heads to be applied to the horizontal and vertical areas of the visor. The yard has stated that it therefore used a Bureau Veritas “Note Documentaire”, number BM2 dated 5.4.1976 for determining the design loads. This note was intended as guidance in the design of the bow of large tankers and bulk carriers. It has not been possible to fully explore how this guidance note was interpreted and used in arriving at the applied loads. The loads so derived were, however, of the same magnitude as those required by some other classification societies at the time.
The design loads to be applied to the attachments of a bow visor of a ro-ro vessel have been continuously developed incorporating new data, and were in general not well established when the ESTONIA was built. The pressure head and the calculation procedure to be applied became more clearly defined and detailed rules were given, for instance, in the 1982 Unified Requirements of IACS and later recommendations. The requirements of IACS 1982 specified equivalent design loads per locking device of about twice as high as those used in the design of the ESTONIA. However, Germanischer Lloyd already in 1978 had a specific formula for the design load of a bow visor which would have given about three times the load used for the ESTONIA.

15.1.2 Shipyard design procedures

From the external design pressure on the visor shell plating determined by the procedure indicated above, the total external load components were calculated by the yard to 536 t (5.3 MN) in upward vertical direction and 381 t (3.7 MN) in aft horizontal direction. These were assumed to act at the centres of the projected areas. The reaction force at the position of the bottom lock was determined by calculations of the momentum about the longitudinal middle point between hinges and side locks at the level of hinges, and found to be 152.5 t (1.5 MN). This horizontal force and the total vertical load reduced by the static weight were divided by 5 and a design force of 100 t (1.0 MN) was obtained as a resultant for each attachment point, hinges included.
Although there is an obvious lack of logic in the procedure used, it was to some extent supported by rules of other classification societies at that time, e.g. Lloyd's Register of Shipping. However, in these design rules, the calculated reaction forces were only to be distributed evenly to the cleats and not to the hinges. It is the opinion of the Commission that the calculations by the shipyard resulted in considerably lower design loads per attachment point than would have been the case if a more realistic design load distribution had been applied.
The design load was used by the yard for calculating a minimum load-carrying cross-section of 6100 mm² for an attachment device. This was obtained by applying a normal stress level of 164 N/mm² calculated from a permissible normal stress of 123 N/mm² for mild steel divided by a material factor of 0.75 due to the intended use of high-tensile strength steel, St52-2. The calculations did not take into account the reduced strength in the shear mode to which many of the attachment elements would be subjected. A copy of the calculations by the ship- yard is included in the Supplement.
The hand calculations by the ship- yard were not submitted to Bureau Veritas for approval. In the actual installation, the calculated effective design cross-section was not incorporated in the bottom lock, nor was high-tensile strength steel used in any of the attachment lugs investigated by the Commission.
The von Tell assembly drawings did, in accordance with the purchase order, identify the operational loads the hull would have to absorb via hinges and operating devices due to the weight and geometry of the visor and the ramp. Loads to be absorbed at the attachment devices due to wave-induced forces were not indicated on these drawings.
Bureau Veritas communicated in March 1980 with the von Tell company regarding the design loads used by von Tell in determining the strength to be built into the locking devices. The von Tell company explained in a brief telex that they had used the rules of Lloyds Register of Shipping in the absence of Bureau Veritas rules and had calculated a load of about 80 t for each device. Although no details are known, the outcome of this correspondence seems to have been satisfactory to Bureau Veritas.
Two notes made by the Bureau Veritas surveyor, one on the assembly drawing of the visor and one on the von Tell general arrangement drawing for the visor and ramp installation, stated that “Arrangement of locking devices subject to the approval of the National Authorities” and “local reinforcement of the ship's structures in way of locking devices, cylinders and hinges to Surveyor's satisfaction.” On the assembly drawing was also a remark that “jack lifting eye on arms, atlantic lock eye, side lock eyes requested in steel grade St52-3” i.e. a high-tensile-strength steel. These drawings were approved by Bureau Veritas with these comments in November 1979 as regards the von Tell drawing and in June 1980 as regards the shipyard drawing.
The shipyard drawing for the visor was submitted to Bureau Veritas for approval only shortly before the vessel was delivered. The note on the von Tell drawing was, however, brought to the attention of the yard by the Bureau Veritas site inspector in March 1980 as recorded in his daily work statement. It is also worthy of note that the von Tell company communicated with the Finnish Maritime Administration in December 1979 about approval of the von Tell design in general but did not then make any reference to the note on the von Tell drawing a month earlier regarding the Bureau Veritas requirement for specific approval of the locking devices by the national administration.
The Finnish Maritime Administration was, under a national decree, exempt from carrying out a hull survey if the vessel had a valid class certificate issued by an authorised classification society. Bureau Veritas did not, on the other hand, make a detailed survey of the visor attachments, as requirements for these were not included in their rules at the time. This situation and the confusing timing of the correspondence about approval of the locking devices, seem to have led to the calculations and the design of the attachment points for the locking devices not being examined for approval either by Bureau Veritas or by the Finnish Maritime Administration.

A visor is subjected to hydrodynamic and hydrostatic loads when the vessel is proceeding in a heavy seaway. Due to the geometry of the visor, the wave load amplitude increases in a non-linear way with respect to the relative vertical motion between the bow and the wave surface. A small increase in the displacement and velocity of the relative motion will cause a significantly higher increase in the wave loads.
In straight head or oblique bow sea, the resultant force on the visor of the ESTONIA would be directed approximately 45 degrees from the waterline due to the shape of the visor, causing upward- and aft-directed load components of equal levels. In bow sea there would in addition be a transverse load component, but mostly smaller. The centre of action of the larger forces on the visor would be positioned high up and forward, causing opening moments, and in a bow sea also twisting and yawing moments about the longitudinal and vertical axes, respectively. The forces from lighter sea impacts would generally cause closing moments on the visor.
Due to uncertainty in the estimate of the sea state, the randomness of relative motions and the non-linearity of the forces created on the ESTONIA's visor, there is considerable uncertainty in the estimates of the maximum loads. On the basis of numerical simulations and model tests (see 12.1 - 12.3) the Commission has concluded that the most probable maximum resultant force on the visor, developing in a significant wave height of about 4 m and after the vessel had changed course at the waypoint, was between 4 and 9 MN. Divided into force components, this equals simultaneous upward and aft forces of 3 to 6 MN and a starboard transverse force of 0.5 to 2.5 MN. The resultant maximum moments about the hinge points were 4 to 20 MNm opening moment, 0.5 to 7.5 MNm twisting moment and 0.5 to 2.5 MNm yawing moment. Load and opening moment levels in the lower part of the range could well have been exceeded a number of times. Levels above the upper limit of the range are judged to have had a low probability of being exceeding, but cannot be excluded.
The Commission has noted that the estimated maximum sea loads at the time of the accident in terms of vertical and longitudinal forces on the visor were of magnitudes about equal to those used by the shipyard as design loads. Later during the accident night the wave height increased and the forces would have increased significantly if the ship had continued at the same speed and heading towards the waves.
The distribution of reaction forces and their directions in the visor attachments are affected by the positions of the attachment points in relation to the position of the centre of wave load action on the visor, the play in the locking devices, the overall stiffness of the visor and the local stiffness at the attachment points. In the following sections the strength of the various attachments is discussed separately based on calculations and tests as described in the Supplement. An estimate of the combined overall capacity is given in 15.10.

The bottom locking device failed in its attachments to the forepeak deck (Figure 8.13). The failure took place by fracture in the three plate lugs carrying the bolt housing and the mating support bushing and in the weld around the housing and the bushing (Figure 8.14). The fracture of the parts indicates tensile failure load directed forwards.
The starboard and centre mounting lugs had essentially failed in the longitudinal plane of the vessel whereas the port lug had restrained the forward movement of the locking bolt and housing and become twisted. The locking bolt had slipped out of the mating lug on the visor at an angle of about 30 degrees.
The failed lugs were recovered from the wreck and have been subjected to metallurgical and strength examination of the fractured surfaces and of the base material (Supplement). All indications are that the lugs failed in a local overload condition with one or a few cycles. The general appearance of the failed lugs is shown in Figure 15.1. The failure in the weld joint was partly in the weld beads and partly in the surrounding material. The thickness of the weld beads was around 3 mm. Indications of pre-accident root cracking or lack of fusion can be seen in the fractured areas of the weld beads.

Figure 15.1 The failed starboard mounting lug for the bottom locking device.

The load-carrying capability of the bottom lock assembly (Figure 15.2) to failure has been estimated using calculations detailed in the Supplement. Only two lugs, symmetrically located one on each side of the visor lug, effectively contributed in carrying load applied in the longitudinal direction via the visor lug. This failure pattern is also supported by the result of tests carried out in Hamburg and referred to below.

Figure 15.2 Bottom locking device.

The load was carried by the rims of the lugs and by the welds between the lugs and the housing and support bushing for the locking bolt. The fracture area of each lug was about 1100 mm² of mild steel, contributing a load-carrying capability of about 0.3 MN to the failure load of the welded assembly. This value is based on a failure mode wherein the lugs would only have been loaded to the yield level when the welds failed due to their lower ductility. The welds had a load-carrying capability of 0.3 to 0.5 MN at each lug, the actual contribution depending upon the quality of the welding and any existence of root cracks.
The total failure strength of the assembly will have been the strength to failure of the welds plus loading of the two lugs to their yield stress, or altogether 0.6 to 0.8 MN per lug. Two contributing lugs would then have given the complete bottom lock assembly a holding capacity of about 1.5 MN including a small contribution from the starboard bracket. The Commission considers this to be a realistic maximum value.
The Commission is aware of a series of tests carried out in 1996 at the Technical University of Hamburg on behalf of the yard with full scale mock-ups of the bottom lock assembly made of high-tensile strenght steel. In these tests, characterised by different extent of welding between the attachment lugs and the locking bolt housing, failures occurred between 1.0 and 2.0 MN. A test that incorporated intermittent welds resulted in failure at 1.42 MN.
The mating lug on the visor had a tensile load-carrying capacity to failure of about 1.8 MN, taking into account the material (mild steel) and that the lug tip was loaded more critically in the shear mode than in the tensile mode assumed in design. The visor lug was therefore just a little stronger than the forepeak deck assembly. Results of analysing the deformation of the recovered lug using modelling and experiments indicate that the lug may, at some time, have been exposed to a tensile load of up to 1.5 MN (Supplement).
The Commission has learnt that the bottom locking device assembly was manufactured by the yard as a shop subassembly that was subsequently welded to the forepeak deck. No detailed drawing with welding data was issued specifically for this subassembly as welding data was generally contained in yard standard tables. The Commission has not found information on any modifications or repairs of the bottom lock. The paint test (see 12.7 and the Supplement) and statements by people involved in maintaining the lock indicate that it is original or dates back to a very early period of the vessel.
To satisfy the outcome of the yard design calculations, the lugs should have had a larger minimum cross-section. It appears that dimensions of the attachment lugs as indicated schematically on a von Tell assembly drawing for the bottom locking device were used in the manufacture of the attachment lugs rather than a design based on the yard calculations. It is noted that the design calculations also assumed high-tensile strength steel, whereas the actual attachment lugs were made from regular mild steel. With regular mild steel the load-carrying effective cross-section of the bottom lock attachments should, according to the shipyard's design calculations, have been about 8300 mm² whereas the built-in equivalent cross-section, including the small welds and the load effectively being carried by two lugs, was only about 4600 mm².
It is concluded that the load-carrying capacity of the installed bottom locking device was not adequate to satisfy the design load and calculated minimum cross-section requirement.

The side locks failed at the attachment of the lugs to the aft bulkhead plating of the visor. No drawing showing details of the installation and welding of these to the visor plating has been identified. An extract from the von Tell assembly drawing for the side locking arrangement was apparently released for manufacture of the side locking lugs. This sketch shows the lugs to have a bottom length of 370 mm, compared to about 550 mm as indicated on the general assembly drawing for the visor.
The lugs were ripped out of the visor plating together with part of the plating itself, leaving rectangular holes, about 390 by 85 mm, with fracture surfaces mainly in shear through the aft plating (Figures 8.19 and 8.20). The lugs remain on the locking bolts in the wreck. The bottom surfaces of the lugs are shown in Figures 8.17 and 8.18.
The thickness of the aft plating of the visor was 8 mm. Two vertical stiffeners were installed behind each lug at the surveyor's request for local reinforcement of the structure bearing the locking devices (Figure 15.3). One of these stiffeners was so located that there was an overlap of the fillet weld of the lug and one of the stiffener weld to the bulkhead plating. The other vertical stiffener had no overlap. The strength contribution from these stiffeners has been estimated to be small. A horizontal stringer on the visor aft bulkhead, located close to the upper corner of the lug, had failed in the plating and partly in the weld between the stringer and the bulkhead surface. No other strength continuity was incorporated behind the lugs.

Figure 15.3 The visor side locking lug assembly with bulkhead plating removed.

The force required to pull and break the lug away from the visor in a direction tangential to the rotation about the hinge points and in the lug plane has been estimated with mock-up testing and calculations to be at most about 1.2 MN on the port side and 1.6 MN on the starboard (Supplement). These values take into account a weld defect at the stringer behind the port lug and the uneven distribution of the fillet weld between the stiffeners and the bulkhead plating. The load-carrying capacity would be lower for a load applied at a smaller angle to the bulkhead plating and higher for a load more normal to the plating.
It is especially noted that the side locks had deficient loading capacity because their geometry induced primary shear in the aft plating of the visor. Thus, although the minimum cross-section of the lugs was almost equal to that required by the design calculation, the strength was, very approximately, only half of what a similar tension-loaded cross-section could provide. However, the cross-sections of the horizontal stringer and the very modest vertical stiffener weld added some holding strength.
The ultimate load-carrying capacity of the side locking lug welds has been calculated to be less than the value above for a weld bead of 8 mm, but the thickness and strength of the weld material is uncertain. The actual failure did not, however, occur in the welds.
It is concluded that the absence of sufficiently detailed manufacturing and installation drawings for the lugs and their supporting structure resulted in an insufficient load-carrying capacity in comparison with the calculated design load requirement.

The hinges at the aft end of the visor deck beams were subjected to loads of about 1.2 to 1.5 MN during normal opening and closing of the visor, acting in directions between downward and aft depending on the position of the visor. The lower rims of the hinge plates had generally failed under tension and the upper ones under bending leaving stretched tongues with strongly contracted tensile, ductile shear fractures of the failed lower rims and flat fractures of the upper rims with clear signs of bending overload (Figures 8.21 - 8.24). The welds to the bushing failed in the forward part leaving the rim attached to the separated part.
The lugs and one recovered hinge bushing have undergone metallurgical and strength investigations as described in 12.7 and in more detail in the Supplement. The investigation of the recovered hinge bushing has revealed extensive cracking in the weld beads, primarily in the downward-facing area. This cracking was initiated by cracks in the root of the weld and progressed through the weld, generally in one of the fusion zones. At a couple of places the cracking proceeded to the outer surface of the welds, as reported after the accident by a student doing paint work on board. The loads generated during normal operation are judged to have been sufficiently high to cause fatigue cracking to progress in the welds, initiated by the original root cracks.
The lug rims ultimately failed in tension of the lower lip and bending of the upper one without previous fatigue cracking, as clearly noticeable on the recovered specimens. The fracture surfaces of the lugs indicate that the final failure took place within one or a few load cycles.
The cracks observed had only marginally influenced the strength of the hinges against wave-induced loads primarily because of the load directions generated.
The load-carrying section of the hinges consisted of the rims of the lugs of the visor hinge beams and the fillet welds around the hinge bushings. The rims of the lugs had a cross-section of 60 by 25 mm in each plate. The lugs of one hinge assembly (two plates) would then, according to simplified assumptions and calculations, have had a load-carrying capability in aft-directed tension of maximum 2.7 MN, using an ultimate tensile strength of 450 N/mm², verified during actual testing. With the minimum cross-sections of the rims at yield in a welded assembly their contribution to ultimate strength in aft-directed tension would be 1.5 MN. In the same loading mode the weld would contribute about 5.8 MN making the total resistance to a single load at most about 7.0 MN for aft-directed tension.
The ultimate strength of a hinge against lifting load has been estimated to about 4.6 MN assuming a clearance larger than about 1 mm between the steel bushing and the lug plate.
The surfaces of the holes in the lugs where the bushings had been inserted had, along much of their peripheries, a very rough contour as from manual flame cutting. This applied to all four lugs, but more so in the starboard hinge. The burn marks were in the lug plates only with no corresponding marks on the recovered bushing. In addition, the forward contours of the holes in the lugs in the starboard side hinge were located about 10 mm further forward relative to the outer contours of the hinge beam than those on the port side. It has not been possible to find the reason for the rough surface, whether it was an adjustment of the hinges during assembly or remains from a later repair. No documentation from any repair in this area has, however, been recorded.
It is most likely that the forces to cause the hinges to fail were generated when the visor, moving upwards around its hinges and having lost support from the locating horns, was exposed to twisting and yawing moments. Given the relatively high stiffness of the deck hinges compared with that of the attachment of the locking devices it is also possible that initial failure of the port hinge was caused by high reaction forces before all locking devices failed.
The Commission notes that the hinge strengths were in general commensurate with design intent. However, the cracks generated during normal service indicate insufficient strength of the hinge lug rims and bearing bushing welds. Also the rigidity of the lug rims to vertically-directed loading is considered poor because of the modest bearing bushing weld and the flame-cut bushing hole with its rather loose tolerance.

The manual locks at each side consisted of two plate lugs welded to the aft bulkhead of the visor and an eye bolt in a cavity in the hull, so arranged that it could be rotated to position between the lugs and tightened. The total load-carrying capacity of the manual locks has been estimated to be at most 0.7 MN each. Had they been applied they may to some extent have contributed to the overall load-carrying capability of the visor locks. The fact that there were no instructions for their use has been taken as an indication, however, that they were not regarded as part of the operational locking system.

The visor had two heavy-duty actuators for controlling the opening and closing of the visor. These were connected to the visor hinge beams at a distance of 1.3 m from the hinges and were mounted on reinforced horizontal platforms in the front structure of the hull. The actuators were connected hydraulically to a solenoid-type control valve, which was closed at all times except when the visor was being moved. Various restrictor valves were installed in the system to limit the speed of opening and closing. The pumps in the hydraulic power supply system had been replaced once with new ones capable of delivering a higher hydraulic pressure as the original ones had marginal capacity.
When sea loads started to open the visor, an upward load was also applied to the actuators, which resisted the opening movement. The leverage from the centre of attack of the sea loads compared to that of the actuators enabled a high pulling force to be transmitted to the actuators. The port side actuator was at this moment pulled out of the hull (Figure 8.26) while only partly extended whilst the locked-in hydraulic fluid acted to transmit the force to the lower attachment of the unit. The vertical force to shear the actuator support out of the hull has been estimated to be from 4 MN down to possibly as low as 2 MN, taking into account the unsymmetrical attachment point of the load and extensive cracking in the platform edges and welds as well as the steel grade used for deck 3. Tests revealed signs of cold brittleness in this steel even at room temperature. The actuator mounting platform has undergone a detailed investigation (Supplement).
The normal operating loads from the actuators appear to have been high enough to initiate fatigue cracking of the platform plating and the welds, in particular where some crack-promoting discontinuities may have existed. The port platform exhibited cracks around a large part of its periphery, generated by vertical loads from normal visor opening and closing.
The seals in the starboard actuator failed, preventing the hydraulic fluid from transmitting the load. The piston rod of this actuator was therefore extended and the actuator remained connected in the hull during the initial phase of the visor movement. The load initially taken is uncertain but must have been below the ultimate strength of the platform, estimated to be below 8 MN.

The ramp was secured in the closed position by six locking devices, i.e. two pull-in hooks at the upper end and two locking cleats along each side of the ramp.
After the accident, the upper pull-in hooks were in closed position as verified by ROV video pictures of the actuator and lever mechanism. It has, however, not been possible to determine in which mode the hooks themselves failed. An upper limit of the load-carrying capability of a hook may have been the load at which the metal in the contacting area between the hook and the mating pin started to yield. This load was approximately 0.2 MN. It is assumed that the pin slipped off the hook when yielding started in the hook material, as the bend-over angle of the hook tip was small.
The side securing bolts, in locked position, extended into box-like structures welded to the side bars of the ramp. These boxes were ripped open following failures in their welds. The force required to rip any one of these boxes open has been estimated to be 0.2 - 0.3 MN. The lower port box, however, was not damaged and it is concluded that the locking bolt was not engaged when the ramp was forced open by the visor. A question remains about the condition of this locking device just before the accident. This did not, however, have any effect on the overall development of the accident.
The locking devices failed sequentially as a result of load applied to the port side first. A force applied to the top of the ramp from contact with the visor had larger leverage than the locking devices had, reducing the force actually required to break the devices. The contact force required to deform the stiffeners in the deck housing of the visor has been estimated at 0.3 - 0.4 MN, sufficient to break open the ramp locking devices.

Other damage to the visor that is related to the accident includes extensive pounding of its bottom and indentations on its front. The bottom plating was forced upwards and had cracks in many places, primarily in welds. The stem post had separated from the side plating and been folded inwards together with the bottom plating (Figure 8.6). Damage marks indicate that this happened when the visor started to tumble forward and was rotating downwards on the ice-breaking prong of the bulbous bow. This damage caused by the ice prong continues upwards along the stem, culminating in large mid-height indentation (Figure 8.5). Further indentations, scratch marks and paint marks on the starboard side of the visor indicate its continued movement when it slid off the bulbous bow and sank underneath the vessel.
Analysis of paint marks in the main indentation shows that these came from paint of the same type as that used below the waterline of the vessel, including the bulbous bow.
It is concluded that the bottom plating of the visor became deformed when the visor was dropping back after having been lifted by waves, initially pounding on the forepeak deck and, secondly and extensively, on the stem head.
Some indications of old cracks have been found in welds, primarily in those between the stem post and the side plating and between the side plating and the bottom plating. Some of these joints may have been exposed to cyclic loading from opening and closing and from waves and ice. Fatigue cracking may have developed, generally starting in stress concentration points in the roots of the welds. It has, however, been difficult to determine the characteristics of the cracked surfaces due to subsequent extensive corrosion.
The stem post, folded inwards under the visor, had four transverse cracks on its front side. It is assumed that these developed during the deformation of the stem by beating on the ice prong although indications in the crack surfaces suggest that cracks may have existed before the accident.
Small paint marks from paint similar to that used on the hull were found in one of the cracks. It could, however, not be established whether these marks were flakes coming from the surrounding paint or had actually entered an existing crack during painting.
It is concluded that some cracks may have developed in some welds during the vessel's lifetime. In view of her age these cracks are considered to be normal and did not contribute to the cause or the development of the accident. Progressive cracking in some weld seams might have affected only the development of the secondary damage, once the damage had been initiated.
Two longitudinal flat bars, though shown on the visor steel drawing as running one on each side of the recess for the locating horn on the bottom plate of the visor, seem not to have been installed. The bottom of the visor therefore had no other structural continuity in its load-carrying members than its aftmost beam to which the visor locking lug was attached. The bottom is therefore considered to have been weaker than intended, in particular when taking vertical loads. This is also likely to have affected the amount of deformation occurring during the accident influencing the ability of the visor bottom structure to resist vertical forces that may have developed during the failure.
The interior of the visor shows several dirty ”waterlines” indicating that water had been standing inside the visor for some time. Some oil, presumably hydraulic oil leaking from the hydraulics of the bottom lock, had floated on top and had settled on the vertical surfaces giving the ”waterlines”. The sealing on the forepeak deck had clearly not always been in a condition to keep the lower part of the visor watertight. The Commission has learned from individuals involved in other ferry operations that this is quite common in many ferries as the seals on the fore peak deck are so easily damaged by the rubbing action occurring during opening and closing of the visor.
The Commission has concluded that the general maintenance standard of the visor was satisfactory. The steel work was little corroded and no reduction of plate thickness, nor pitting, has been noted on the various parts collected for detailed investigation.

The failure pattern of all the attachments indicates an overload caused by forward-upward motion of the visor. The Commission has considered different possibilities that could have caused the attachments to fail, but has come to the conclusion that it was an external wave impact on the visor that created the necessary failure loads.
The visor of the ESTONIA was not fully watertight, and probably some water penetrated into it in the rough head and bow sea the vessel encountered. Hydrostatic pressure from trapped water inside the visor would create a resultant force directed about 45 degrees forward and down. The pressure and the resultant force would be amplified by the vertical accelerations of the bow. However, the possible amount of trapped water could not have created tension reaction forces in the attachments sufficiently high to make any of them fail. As an example, 3 m of water inside the visor would create a hydrostatic resultant force of only about 0.5 MN.
Green water on deck could be critical due to the unfavourable lever arm to the aft-positioned visor hinges. About one metre of water on the deck would double the weight of the visor, but several times this height would be needed to break the attachments. The model tests and the numerical simulations show that the probability of any significant amount of green water on deck was negligible for sea conditions at the time of the accident.
In general, the lower level of wave-induced forces on the visor occurring every few minutes in the prevailing sea condition would cause closing moments about the deck hinges. The reaction forces so created were taken up by the stem post and steel pads on the forepeak deck. The model test results indicate that the maximum closing moment including the visor's weight was about 8.0 MNm. If the stem post alone had supported the visor, the stresses from this moment would have been about 120 N/mm² in compression, possibly with some added bending stresses. If the visor locking devices had taken up the closing moment, the reaction forces would have been directed forward, causing compression in the lugs. The closing mode is not considered critical for the attachment system.
When the resultant wave-induced forces exceeded 2.0 - 2.5 MN, the line of action would pass above the hinge axis and create an opening moment. With an average period of about 10 minutes in the accident condition, the opening moments were high enough, about 3 MNm, to exceed the static weight of the visor. Less frequent, higher wave loads caused much larger opening moments. Opening moments would create tension in the visor lugs of the locking devices and forward-down compression in the visor hinge lugs. In oblique bow sea the reaction forces would be unevenly distributed between the side lugs, increasing the tension on the side that encountered the waves. This effect is clearly visible in the visor damage which shows strong movement from port to starboard.
The attachment system was statically undetermined and the distribution of reaction forces was therefore affected by the stiffness of the structure as well as by the play in the locks. By studying different levels and combinations of wave load components and different possible distribution of reaction forces, an estimate of the total load-carrying capacity before failure of any of the attachments has been obtained. The estimated strength of the individual attachments is described earlier in this chapter. Assuming that all locks worked efficiently and using realistic correlations of wave load forces and moments obtained from model tests in port bow seas, it is estimated that the combined strength of the attachment system would be exceeded by an external resultant wave load of 7 - 9 MN, corresponding to opening moments 13 - 20 MNm.
Most likely the port side lock failed first, possibly at a lower load level than the maximum estimated above. The subsequent failure could have occurred either by shearing of the port hinge lugs, or by tension in the bottom lock. The estimated necessary level of wave load for either of these possible second failure modes is about the same. Hence, the complete failure sequence could have required only one or two wave impacts. If only the port lock failed in a first large wave impact and the other attachments remained intact, then the visor was probably kept in place - appearing undamaged - for a significant period. The hypothesis of a side lock failing first is supported by the similar damage pattern incurred by the visor attachments of DIANA II in 1993.
Figure 15.4 illustrates an example of a possible reaction force distribution over the attachments when the port side lock fails. The load on the hinges, though large, is acting in an uncritical direction while the bottom lock and the starboard side lock are loaded only to about half of the critical level. A possible reaction force distribution after the port side lock has failed is shown in Figure 15.5.

Figure 15.4 Example of reaction force distribution resulting in port side lock failure.

Figure 15.5 Example of reaction force distribution after the port side lock had failed.

The final stage of the attachment failures leading to the loss of the visor took place in the remaining hinge lugs and in the lifting actuator sealings and mounting platforms when the visor was free to move. Significant forces in the actuators are judged to have developed only after the locks had failed. The maximum necessary resultant wave load to fail the actuators simultaneously would be about 6 MN, but they most likely failed in sequence under dynamic conditions at a significantly lower level.
The extreme value distribution of wave-induced loads given in 12.3 for the accident condition indicates a theoretical probability of more than 5% that a single wave load would exceed the largest estimated combined strength of the attachments during 30 minutes at a speed of about 14 knots in oblique bow seas with a significant wave height of 4.0 - 4.1 m. The probability of wave loads exceeding the attachment's strength increases rapidly with increasing wave heights.
The Commission concludes that the largest forces generated by sea loads under the prevailing conditions were higher than the combined strength of the attachments and hence caused the failure of these and the subsequent loss of the visor.
It is notable that the ultimate strength of the visor attachment system was exceeded already for a load level about equal to the design load used, and that this load level developed in a sea condition that was far from the worst the ship could have been expected to encounter. There was consequently no margin of safety incorporated in the visor attachment system.

After having studied the design, manufacturing and procedures for approval, the Commission finds that none of the parties involved considered the visor attachments as critical components for the safety of the ship. There is however no indication that the routines in this respect deviated in general from routines followed by other parties for other newbuildings in the Baltic area at the time. Information has been found, however, that in some other parts of the world in the 1970s visors and their attachments were more thoroughly designed and considered critical for the vessel's safety.
The ESTONIA was designed and built after a decade of very rapid development in shipbuilding and naval architecture. Ship sizes increased, shipyard technology was modernised and new computer-based direct methods for calculation of structural strength and wave-induced loads were introduced in the design process. As a consequence, experience lagged behind technological development. Ships are in general highly optimised with regard to structural strength, and experience from damage, incidents and accidents has always been an important basis for the development of codes and procedures.
Ro-ro passenger ferries for the Baltic traffic developed very fast during the 1970s as described in Chapter 10. The ESTONIA was at the time of newbuilding among the largest bow visor ferries ever designed, and experience from similar designs was obviously limited.
Today with the outcome known, it is easy to find several items to criticise in the design of the visor. However, if anyone had made a rational analysis of the concept, the same items could have been found open to criticism even at the time the ship was designed. Here are listed a few such design considerations that the Commission regards as important.
All attachment points for a bow door should be regarded as highly loaded design items and subjected to detailed load and strength analysis.
Where operating experience is limited and design rules and recommendations give little support, it is of utmost importance to make a failure consequence analysis. Even a very simple analysis would in this case have highlighted the critical interconnection between visor and ramp and the possible consequence of water on the car deck after failure of the visor attachments. The conclusion would then have been either to separate the two barriers or to incorporate a very large safety margin in the design of the attachments.
The design load calculation and the assumption of equal distribution of forces on the attachment lugs for which there was also support from some classification societies' design rules at that time, had no basis in physics. The reaction forces obtained were not in balance with the external load, and no specific directions of forces were determined. Since the design pressure head was equally distributed over the shell of the visor, oblique loading was not even considered. The Commission is of the opinion that even very simplified design calculations of vital items should include analysis of different possible load directions and failure modes. In a statically undetermined system of supporting points, either detailed analysis including flexibility should be performed or sufficient strength should be assessed for any combination of reduced system that could be evaluated by simple equilibrium of forces.
Locking devices of the common design such as those installed on the ESTONIA are subject to wear and corrosion on the locking bolts and mating lugs, leading to play in the arrangement. Some initial play is also built in, to safeguard functioning during opening and closing of the devices. The play was about 10 mm in the ESTONIA's locking devices and extensive play of about 35 mm is known from other ships. Play between the connecting parts has the consequence that the contact load distribution between the locking devices is undeterminable and in an extreme condition one single device may be subjected to the entire external load. Play in a connection subjected to dynamic loads will always lead to accelerated wear and may induce fatigue. The Commission is therefore of the opinion that locking devices should be of such design that play is eliminated during closing and bow doors should, in the closed position, be physically tightened against mating surfaces.
The local design of the ESTONIA's visor attachments shows weakness particularly because the lowered strength in the shearing mode had been ignored. The hinges were weak due to weld shear and bending of lug rims induced by vertically directed reaction. The side lock induced shear in the visor aft plating in all load modes, and the bottom lock design capacity would have been limited by shear of the visor lug tip even if the forepeak assembly had been welded to better standards.

The design requirements for the visor attachment devices at the time the vessel was built generally indicated a design load level of about 1 MN per device based on an even distribution of the bow load on all the attachment points. This applied to the way the loads were determined by the yard, and also followed from the rules of some other classification societies at that time.
The ultimate failure load, corresponding to the 1 MN design load, would in the simplified analysis made by the yard have been about 3 MN per device taken as the ratio between ultimate strength and permitted stress. However, in a mixed design where various design elements are included and different failure modes will develop sequentially, this relationship is not directly applicable.
The failure load of the bottom and side locking devices as installed has been determined to be about 1.5 MN per device, account taken of the uncertainty in load direction and failure modes.
It seems obvious to the Commission that the ultimate load-carrying capability of the attachment system and its individual devices would have been considerably higher if the reduced strength in the shear mode had been considered and the manufacture would have been according to the design intent.
The design loads used were, however, low and it is noteworthy that even an installation fully meeting the design assumptions applicable at the time of construction would not in all cases have withstood the hydrodynamic forces generated during the night of the accident, considering the increasing wave heights. As an illustration, the model test results show that the maximum opening moment generated by wave loads would be more than three times higher than estimated for the accident condition in bow seas with a significant wave height of 5.3 m and a speed of 10 knots.

A number of serious incidents are referred to in Chapter 11. The cases mentioned are those for which the Commission has found information without extensive research. The ships involved were in most cases operating in Scandinavian waters. It is fair to assume that a number of incidents have taken place also in other trading areas.
In the individual ship involved in an incident, the relevant structures were in most cases reinforced during the necessary repair. In several cases reinforcements were also made on sister ships. However, in some cases the classification society involved was satisfied with a repair restoring the strength of the device to the original standard. This was the case e.g. after the DIANA II incident in January 1993. The vessel was at that time 13 years old and the damage seems to have been regarded as isolated and no further alarms were raised or action taken.
The Commission has noted that in several cases the affected Administration communicated with the classification society involved and was satisfied with the information that the strength requirements had been increased. This would apply, however, to new ships only. Retroactive upgrading of existing ships through rules and regulations with retroactive effect has generally been considered unacceptable within the shipping industry and this attitude has been accepted within IMO as well as amongst classification societies.
The Commission finds this attitude unacceptable in cases of incidents with serious safety implications. The Commission is of the opinion that all amendments to requirements, founded on actual risks having become known, should lead to requirements with retroactive effect, both in IMO regulations and in the rules of the classification societies. The Commission has also noted that a move in that direction has taken place since the ESTONIA accident.