DESIGN CRITERIA FOR WEATHER ROUTED TRANSPORT A.B. Aalbers (Maritime Research Institute Netherlands, MAR1N), R. Nataraja (Noble Denton Europe Ltd.), S. Anink (Dockwise Transport B.V.) SUMMARY The paper describes the advanced design process for weather routed heavy cargo transports which is now possible using the Safetrans PC software tooi. The software tooi, devised as a risk based, probabilistic response method for the calculation of the design loads for ocean transports and tows has been subjected to a thorough calibration and validation. To qualify as an engineering method, a load factor study was carried out to estabÜsh partial safety factors as function of probability of failure. The target level of safety (i.e. upper level of acceptable probability of failure) depends on the type of transport and has been identified in a consequence class evaluation. This allows the user to select the appropriate safety factors for design. An engineering guideline has been developed to instruct the engineer as to selection of targets and proper use of the method. The paper reveals the various specific capabilities of Safetrans and presents a sample case of results, illustrating the effect of weather routing. 1. 1NTRODUCTION capabilities. 1.1 BACKGROUND Safetrans is the only design tooi available to model weather routed transports and offshore operations with weather windows. The system uses a historie weather database that includes the weather forecast. Hence, the simulated transport will not leave when bad weather is predicted or will go around bad weather or go for safe haven. For weather-routed trips the maximum allowable condition such as leg bending moment, cribbing pressure etc. is known in advance. From that, a limit sea state can be derived. The Safetrans software tooi has been developed in 1998-2001 in a Joint Industry Project with Oil Companies (11), Heavy lift transport and towing cöntractors (7), Warranty surveyors (4), Class Societies (2), Drilling cöntractors (2) and Engineering companies (6). The co-operation in this broad field of interest and ski lis ensured tight control of the enabling technologies for the risk based, probabilistic design methodology: MetOcean physics Risk Analysis Ship hydrodynamics Long term statistical evaluation techniques Ship routing logies & experience The completion of the software was presented to the public in September 2001 at the City University JackUp Conference [1]. Since then, the User Group has taken over responsibility for further development and user support of the software. A major part of this development work is the Engineering Guideline and Load Factor study presented in this paper. 1.2 SUMMARY DESCRIPTION SAFETRANS The Safetrans software tooi is a Monte-Carlo Simulation method: a randomly repeated weather routed voyage simulation. From the departure location to the destination the voyage is simulated in 3 hr time steps. In each step the ship 'master' (Captain's Decision Mi mie) decides how to proceed, based on forthcoming information on weather forecast, sea piloting area, actual and forecasted ship behaviour and on ship Figure I: Safetrans Monte Carlo process The simulations start at random departure dates, chosen in the season of interest. During the trip, the response to each sea state is taken into account to obtain a long term distribution of the vessel response over the voyage. By repeating this exercise sufficiently many times, the ensemble long-term distribution of the motions is obtained, which can be used for design. The Monte Carlo Simulation method allows the user to determine the operational criterions under which the transport can be carried out to an acceptable risk or delay. Presence of tropical cyclones is taken into account, and if a simulation accidentally hits a tropical cyclone the simulation time step is reduced to 1 hr in order to account for the rapidly changing weather conditions. Compared to conventtonal design methods, using design waves and assumed worst heading scenarios the Safetrans method is distinctive in the foliowing: ADVANTAGES; Swell, wind force and waves combinations are correctly modeled and included in e.g. load calculations. Bad weather avoidance and bad weather tactics are modeled, ensuring realistic vessel heading with respect to the environment Short trip scenarios with use of favourable weather windows can be modeled. Offshore operations with installation criterions can be modeled. (Float Over) Failure risk assessment is possible, allo wing the user to design to a pre-defined reliability goal. Realistic estimate of voyage duration is possible A quantitative risk analysis (economie, personnel and environmental) is carried out. Fatigue analyses are possible on basis of the logged calculation results. DIS ADVANTAGES: More operational and technical knowledge required to understand the calculations More time consuming than design wave method 1.3 1.3.a is the structural capacity of the cargo, others from operational considerations like clearance of overhanging cargo, t h e wave height and wind speed criteria are default requirements, and additionally the tow force criterion for tows. A total of eight criteria may be specified. The criterion input is in the form of an 'operational criterion value' and a 'safety factor', see Fig. 2. KUnBQXD!££flfl~ Ofcrin *dM -1 £(mttfrf«ja te W| nraUcm roü <XG )dtq{T • aata nottan VX frifl wjtton pltoA 015 tcfajjl pffll •UtffFnJu Ml 41.1 6.1 t l . Il s.nf ï.a l.S 1.3 1.5 Bm.» 1 Bit Figure 2: Criteria input Additionally, criteria that define when the CDM has to consider change of power & heading for comfort can also be given. The operational value is the value that the CDM has to try to avoid by re-routing, sheltering and changes of course and speed for comfort. The safety factor is used to determine the maximum acceptable value for cargo, sea-fastening-or ship.- If this value is exceeded the MCS risk analysis will assume that the probability of damage is orders of magnitude larger than otherwise. The operational values are used to calculate a weighted "ship status variable" upon which the CDM reacts. The operational criteria for motions, etc. depend on cargo and ship limitations. It is advisable to make a consistent set. For example, for roll and pitch: criteria for both are listed because these motions are sea direction dependent and roll is generally low when pitch is large and vice versa. Both criteria have to be consistent with the wave height criterion, which can be achieved by using motion database results as sketched below. DESIGN USE Criteria The Safetrans software is focused on a probabilistic design format instead of the prescriptive, simplified format commonly used for design of marine operations. Even the new ISO code (ISO 19901-6, 2004) is based on prescriptive values of motion responses and allowable friction coefficients etc. A probabilistic, risk based method needs criteria which define what a transport can safely accept and thereupon compute the risk. Additionally, the Captain's Decision Mimic needs criterion input to safely 'steer' the vessel over the seas. Some of these criteria are derived from Roll criterion in Scatter diagram Figure 3: Wave height criterion and motions can be related as shown The values of the criteria have to be realistic. For example, it is not advised to try to cross the North Atlantic Ocean on a northerly great circle route with an operational criterion for the significant wave height of only 3.5 m. In winter half year the vessel will not be able to properly re-route and extreme voyage durations will result: actually the method will generally make the vessel wait until summer. 1.3.b Risk The risk calculation is accumulated over the 3 hrs time step process. In each time step the actual conditions, e.g. waves, wind, vessel motions, tow forces, etc derived from the MetOcean database are accounted. The table below reviews the risk calculation. Tabte 1: Risk Hazard (Initial event) Capsize Conditional Collision *) Fire/Explosion *) Foundering Historie Historie Conditional Grounding (powered) Ship stability failure Machinery Failure *) LossofControl *) Conditional Structural Failure Sea Fastening Failure Conditional Conditional Towline Breakage *) Conditional Towline Fouled *) Other events *) Secondary grounding *)Delay due to Historie drift 1.3.c Historie Historie Event count Evaluation in Risk module Description The historie capsize probability is enhanced by a factor 10J when the Capsize Risk Decision Variable (DV) exceeds a criterion value. This DV is large when e.g. resonant roll occurs and/or when relative motions at the side exceed the freeboard. The historie average collision probability is used The historie fire/explosion probability is used The historie foundering probability is enhanced by a factor 10J when the "Green water on deck" Decision Variable exceeds a criterion value. This DV is large when relative motions at the bow and the sides exceed the freeboard. The historie grounding probability is enhanced by a factor 10 if the vessel is close to the shore (< 1 nm) The historie probability of a ship loosing ïts stability due to a damage (leak) or shifted cargo is used The historie machinery failure probability is used Count of the number of times that the tow line maximum force in a 3 hr MC simulation step exceeded the break strength. The historie structural failure probability is enhanced by a factor 5.102.(R)2 for R>0.5. The value of R is the average ration of SDA value of the criterion signals and the design limits (which are given by operational criterion times safety factor). The historie tow line breaking probability is enhanced by a factor 103 when the 'Tow line break risk" Decision Variable exceeds a criterion value. This DV is large when the most probable maximum tow line forces exceed the operational criterion times the safety factor (= tow line break load). The historie tow line fouling probability is used Historie The historie probability is used Both these secondary hazards are computed on basis of historie average recovery times for the given initial events (or tug assistance if the vessel is incapabie of self recovery) and the drift time to down-wind shore (using wind speed, vessel drag and distance) 10 Voyage return value for design The computed results of risk, ship motions, most probable maximum values in each time step, are logged. For design purposes it is necessary to carry out multiple voyage simulations (typically >200, but 250 in the Load Factor study). The software carries out an ensemble analysis for multiple voyages, in which the results are statistically evaluated to obtain the 10voyage return values. In a sensitivity study on the various possibilities for ensemble long term statistical evaluation, it was shown that the P90 value, i.e. the 90% nonexceedance value of the individual voyage's Long Term Most Probable Maximum values, was the most adequate definition of the characteristic value for design of weather routed transport and hence 10voyage return values. The definition is considered consistent with experience in transport history. 1.4 ENGINEERING GUIDELINE FACTOR STUDY & LOAD In 1999, a load factor study was carried out for the statistical design calculation process denoted Voyage Acceleration Climate (V.A.C.). This study is referred to as the 'First Load Factor study' carried out in the Safetrans TOW JIP [1,2]. The V.A.C, or renamed Voyage Motion Climate method (VMC) in Safetrans is based on probabilistic, response based statistics for the motions on a route, given the scatter diagrams for sea climate. The method has upon completion of the load factor study been accepted as engineering method for (heavy lift) sea transport. The present load factor study and the resulting engineering guideline enhance the present knowledge and is expected attain industry acceptance for future designs. The safety level implied by using a probabilistic method like Safetrans depends on the definition of the characteristic values and safety factors. The Engineering Guideline & Load Factor Study has been carried out to obtain a proper definition for the characteristic values and to establish the safety factors to be applied. In particular it is crucial how the load effects on cargo and sea-fastening are defined in view of the period (of the voyage) that is considered. The definition of the P90 value, together with the systematic uncertainty analysis for the calculation method and environmental data has determined the bias of the load effects. A particular issue to consider in this context is the structural interaction between the cargo and the vessel when it is significantly large (e.g. a production platform). This interaction depends upon the relative stiffness of the cargo and vessel and on friction effects in the cribbing layer between the ship and cargo. To ensure a consistent definition of 'characteristic values', i.e. the above mentioned P90 value, the calculations with Safetrans should be carried out properly. To that purpose an engineering guideline has been developed which instructs the user to correctly model the voyage in Safetrans. For the definition of the safety factors, which have to be applied to the P90 characteristic values to obtain the design values, a partial safety factor calibration study ('load factor study') was carried out. The load factor study is described in the next section. In the sea fastening example, it comprised systematic uncertainty analysis and takes into account friction in the cribbing. The elastic properties and layout of the cribbing are assumed to un-couple the structural stiffness of cargo and ship. The load factor study resulted in safety factors as a function of probabiHty of failure. A consequence class assessment was carried out to define for which types of transport which level of failure probabiHty would be acceptable. 2. SAFETRANS PARTIAL SAFETV FACTOR CALIBRATION 2,1 CASES . . - The partial safety factor calibration study has been carried out by Noble Denton Europe with input from heavy lift transport contractors and Load Factor Study Working Group. Table 2 shows the study cases which were made available. A selection was made on basis of comparability with and availability of original design data and relevant signal analysis results. Table 2: Transports evaluated for Load Factor study Case VESSEL START PORT CARGO END PORT 1 Transshelf Galaxy II Singapore Hal i fax 2 Blue Marlin Glomar Adriatic Camel on Alexandria 3 Mega Marlin Thunderhorse Semi Korea Gulf of Mexico 4* Jumbo Spirit 400 t Columns Japan Persian Gulf 5** Giant Barge Bridge Girders Cadiz Malmo 6 Happy Buccaneer 10001 Ship Loader Brisbane Dalrymple Bay 7*** Boa Barge 10 Maersk Rig 62 Brownsville Maracaibo 8 Tai An Kou Sea Star Singapore Corpus Christi * ** *** Not included in calibration - small cargo Not included in calibration - CG accelerations not supplied Not included in calibration - Original design forces not supplied. The selected cases had the following propeities and actual (historie) design values: Table 3: Transport data 2.2 (0 Design Force (kN) Yield Stress (MPa) Cargo Value <M$) Mass Selected Cases Transshei f/Gal axy 24040 44204 235 10-100 Blue Marlin/Adriatic Tai An Kou/SeaStar Mega Marlin/Thunder Horse 10261 235 5839 60000 41975 33347 147150 235 235 10-100 10-100 Buccaneer/lOOt Ship Loader 1030 4345 235 LOAD FACTOR STUDY: BACKGROUND The cases are evaluated based on total transverse load, represented by the transverse wind load plus the inertia force due to the transverse acceleration at the centre of gravity of the cargo. When this load (Ddemand) exceeds resistance (combined effect of sea fastening (Fdeslg^ and friction), that will lead to total failure (G4>) where the G is the failure function, also called the "G function": G = | k.Fdes(gn +C/. M.g\- The code calibration (systematic accuracy analysis) is based on results from the first load factor study for VMC as well as the new results from an investigation of the weather databases and statistical reliability of multiple voyage simulations [3, 4] c) This leads to a consistent set of Type I (basic physical) and Type II (model ing) accuracy distribution functions, The failure function can then be described as: G = VnvF}a„0Jy Figure 4: cribbing friction distribution Ddemam/ b) e) From the histogram of Most Probable Maximum values for the voyage (an example is given in Fig. 5) the P50 and P90 values are derived, which two values are sufficiënt to fit a 2-parameter Gumbel asymptotic function to the tail of the histogram. The mean and Standard deviation of this Gumbel function is used in the Load Factor evaluation. So, a realistic probability that higher loads than the P90 occur is taken into account in the calculation of failure probability. +lfCrr]cMg\-r]cT]mTis7]hD in which the Type II uncertainties (normal distri butions) are given by the following bias and Coëfficiënt of Variations: r)w: Welding uncertainties u = 0.8, C o V = 2 0 % T)f: Friction (factor) uncertainties H=1.73,CoV=29% nc. Mass uncertainties (lognormal) >i=1.05,CoV = 5 % lm, tls, T|h- Analysis uncertainties li=1.0, 1.0, 1.1 CoV= 5%, 5%, 10% d) 100-400 consistent with the practice of e.g. heavy transport operator Dockwise. Note that the accuracy distribution applied on it (see Figure 4) results in a 95% probability that the friction is between 11% and 41%. A simplified overview of the methodology for code calibration and assessment of optimised partial load factors is given because the theory and methodology is quite complex. The following aspects are important: a) >400 An average friction coëfficiënt of 0.15 for the cribbing was used. This is conservative but 30 25 / |20 l 15 10 / n 1 1 2 3 * s N II II n TT ftn. 5 6 7 8 9 10 11 12 13 W 15 16 17 18 19 20 21 22 23 24 Amplitude Figure 5: Histogram of Most Maximum values of Multiple MCS Probable With all the accuracy distribution functions in place, the probabiHty of failure can be computed for a given safety factor k, with F(tefiff, S( Ddemand FfHojonJ/k. This is done by the software package COMREL [5]. A 3rd order function was fltted through the numerical results yielding p = f(k), with P being the safety index. If OK, these partial load factors could be considered as optimised and be used in other transport designs. Fig. 7 shows that good consistency was obtained: the revised reliability spreads only lightly around the target value of 2.942, which was the historie average of the 5 selected transport cases. Bata va Factor ol l a f a l y r *! • g . B t l B l ' . l . f l I i ' * 1.417». * I.MIZ R'"1 ^ — — Avaraga ^ — P o l y . (Avaraoa) ^*>*~~^ ^ ^ O.S 1 1.1 2 1S Faclor ef Safaty Figure 7: After calibration and optimization the revised reliabilities show little spread Figure 6: Fit function for safety index Then, as shown in the equation below, partial load factors YR, YD and YF were attributed to FdMign, Ddemand & n d ^resistance'^design —"•0 / dcmand - ^' — ?fric«ion '*-f ' ^ 'S) A sensitivity study showed that Yfnctioo could be fixed to a value of 0.7, leaving only two partial safety factors to be evaluated As a result, p becomes a function of k which depends on YR and YD^ Note: of the 5 cases considered, 2 were very heavy lift cargoes and 3 were lighter in comparison. Initially, some concern was present whether these cases could be compared, but in the evaluation of partial load factors the consistency of the results was quite good and there was no systematic difference between the 3 Üghter and 2 heavy cargoes. The following overall result was obtained: k = (-CfM.g + yD.D)/yR. Fdeslgn o Hence, for a given YR and YD it is possible to calculate p. Hence, for each transport case and each MCS type a 2-d matrix of P's was constructed. The partial load factor evaluation is then carried out, which is the minimisation of the error in comparison with the target safety index, i.e.: Error = £ ( £ « , « - f l ) 2 In this minimisation YR and YD are varied and that combination of YR and YD are selected where Error is as close to zero possible for each selected target value (i.e. resulting in a probability of failure of 0.1,0.01, etc). g) 2.2 In this minimisation of Error, the summation includes every transport case for which an actual design value was available and which was considered not too different from the average transport type conditions. With these YR and YD the probability of failure Pp of the cases was recalculated. LOAD FACTOR STUDY: RESULTS The PF value was compared with the target and inspected as to consistency between the cases. Partial Safety Factors Case Pf PfO.1 1.G0EO1 W C A/Bete 2 6 9 6 0 2 PT 0.01 1.C0EO2 fin Bèta PT 0,001 FfQOOOI 1.63603 1.00EC3 1.00EO4 P IresstaTce idEmand Tfricticn 1.282 070 070 070 077 077 091 070 088 1.12 1.5* 1.65 1.80 070 070 070 070 O70 070 1.928 2326 2942 aooo 1719 Table 4: Partial safety factors as function of probability of failure The case descriptions in above table give the target probability of failure, where the present industry Standard is represented by the case VMC Av. Bèta. In the first Load Factor study this average reliability was established for the Voyage Motion Climate calculation method. Additionally, and in a similar way load factors have been established for component design, where cribbing friction is not present. Safetrans allows computation of component loads, e.g. leg bending moments for Jack-Up rigs or internat structural loads in container cranes. Thereto the user can define linear combinations of motions and mass coefficients. 3. CONSEQUENCE CLASSES AND TARGET SAFETY 3.1 CONSEQUENCE CLASSES criteria are adopted. For control of crew risk there is the ISM guideline, but for structural safety the industry itself has to set the target safety (probability of failure). The Engineering Guideline of Safetrans provides the table below for the grouping of typical transport cases in various 'consequence classes'. Various questions may be raised as to consequence evaluation, and the most important will be discussed. The concept of consequence classes basically addresses the marine safety regime. The present convention is that for environmental and human risk, the IMO acceptance Table 5: Consequence class definition C o m i a u a nc« C l a a i CO Slmpla oparallon High Radundancv C 1 W «II controllad o p a r a t l o n with high r a d u n d a n c y C 2 C o m p l a i or w « a t r i a r canaltlve oDeratlon C3 C o m p l t i and waathar • i n attlva o p e ra tlo n T y p » of o p a r a t l o n O a n a rat c a r g o u p to 2 0 0 t o n i C • n taln a r i R oirina S t o c k D ry t r a n i p o r l l o w v a l u s • Jack Up» ( * m a l l m « d l u m ) Type O l a n a l v i l i C argo aaeurlng m a n u • 1 w llh d a a l g n m o l i o n d la d r a m t - Birgti 0 p a r a Ilo n a 1 r a q u Iro m a n t t Saa VM C - D radge Cargo - C o n taln e r c r a n t i • Dry t r a n a p o r l H i g h V a l u a • W ealhar roulad Iranapori Low V a l u a W * a t h a r roulad I r a n i p o r t high va lu a W « l tow Of h i g h v a l u a VMC IM 0 S a a IM 0 or M C S On board axpert lyatam ( 0 B A S ) o p a i a t a d by MCS On board a ' p a r t i y i t a m w Ith o n b o a r d e x p a r t MCS Human safety: The IMO guidelines address the role of human safety in the consequence class definition and in the target criteria setting. In reality these guidelines were developed for passenger craft and any such marine craft where total loss is synonymous with significant human fatality. Against this background, in the case of heavy lift transports: • Most of the transports have un-manned cargo with minimum crew on board the transport vessel • The heavy cargoes have an asset value running into 100sofmillions(ofUS$) • The economie value of loss of cargo, in addition, is orders of magnitude larger than the asset value. Therefore it is unrealisttc to contemplate IMO human safety criterion as a measure of consequence class for heavy lift transport. Nevertheless, the Safetrans risk analysis facilitates computation of human fatality risk and other hazards for the ship and cargo should this be a requirement for significantly manned transports. It should also be noted that Safetrans takes due account of the ship crew ISM rating. operators and warranty surveyors (e.g. Dockwise, Jumbo, etc, and NDE and MatDan) confirm that in many sea-fastening designs there is redundancy. Therefore, the presence of redundancy is proposed to be a determining factor in the consequence class grouping as given in the table above. The Safetrans Quantitative Risk Analysis shows that searfastening failure is only. one of the 7-9 hazards identified to possibly lead to failure of the ship and/or cargo on the voyage. However, for cargo safety the proposed design approach fulfiïls the objective. Effect of fatigue: Sea-fastening is usually designed not to attract fatigue loading. Cargo owners, however, have to consider fatigue contribution to the cargo structure during transport, for which Safetrans facilitates output of fatigue load cycle distributions. 3.2 TARGET SAFETY The plot in Fig. 8 is the final result of the Load factor study and shows the relation between -yR and yD (with YP = 0.7) and the probability of failure. On basis of a prescribed probability of failure for a given consequence class the partial safety factors can be read off. Does component failure lead to total failure? The Load Factor Study assumes total load on the sea-fastening and in case of failure a total loss of cargo. Transport PARTIAL LOAD FACTORS 2.0 m I I , . | I l *~ Kesistance ~"~ Demand '"—^ I I 1.0 — — _ , —L_ J. — ^ • *i-j I i i i 1.00E-04 1.00E-03 i 1 1.00E-02 Probability of Failure Figure 8: Load factors for weather routed transport design using MCS \ 1.00 E-01 1 1 1 1 I t 1lÉ1 um mi . • 100E-04 1 1 1 1 Revised Force/ Original Force Pau U B Prctebdity a 1.O0EÖ1 •269EC2 Q1.0CE02 C31.63E03 •1.00EO3 Figure 9: Design loadsfor selected Load Factor cases 4. DISCUSSION OF RESULTS FOR LOAD FACTOR CASES 4.I NEW DESIGN VALUES The new design loads are computed using the partial factor for demand from this table together with the P90 acceleration and compared with the original design load. The results are given in Fig. 9 above for a range of probabilities of failure. An interesting conclusion is found after comparing the design value from the Safetrans MCS method with the existing designs for the 5 selected cases. It shows that according to the MCS calculations that reflect the experience of actual transports as they occurred, these transports are, on average, an order of magnitude safer than it was perceived from the existing design method. Hlndcasl 11% SWÜstlcal ---— 6% ^ Modeüina J^^L. e* Cargo Mass r^^ ^ ^ ^ ^ ^ ^ frictlonX. / 18% \ ^ wair ModBlling 4 % ^ 4% ^ / """^^ ^ ^ Cargo Mass i ^ ' ^ f c ^M 1 ^ Another question may be the comparability of the transport cases. It was discussed before that the results of the load factor optimization were -consistent. It is-interesting though to see in Figures 10a and 10b the typical sensitivity plots of the contributions of the various uncertainty parameters. It may be clear that for very heavy cargoes the friction is the major single factor. Hlndcasl Capaclty Statistica) 8% 5% Capaclty 7% " wm ^T One may wonder if this may suggest that the P90 value is too optimistic. This is not the case because the code calibration uses the Gumbel asymptotic fit to the histogram of voyage maxima (see Figure 2 above). Therefore, realistic probabilities that loads occur which are higher than the P90 design value are included in the calculated failure probability. So, the conclusion is justified. Further this conservatism is also justified in the light of consequence of cargo loss. • Trans.Ace ' 40% ^ 8% Sensitivity Plot for Average of BOA, Adriatic& Sea Star r ± m è^_^/ k Trans A c e 25% 7wdd / 5% Sensitivity Plot for Average ofThunderhorse and Calaxy Figure 10 a and 10b: Difference of cribbingfriction effect between heavy and light cargoes 4.2 CASE: GLOMAR ADRIATIC TRANSPORT The 10265 t Adriatic Jack Up rig was transported from the Gulf of Mexico to the Mediterranean Sea on the Blue Marlin of Dockwise. The voyage was planned for January-February departure. Since the legs were fully erected, the leg bending moment was the limiting criterion for weather routing. On basis of Dockwise response calculations the snip had to avoid 6.7 m beam seas and avoid head seas or bow quartering seas exceeding 7.6 m Hs. In Table 5 the transport would be Consequence Class C2. The actual design load for the sea fastening (including wind effect) was based on design wave response calculations, which were quite comparable with Voyage Motion Calculations (VMC) shown in Table 6. On basis of the First Load Factor study, the VMC method is accepted industry Standard. An operational criterion of 6 m Hs for weather routing and shelter options (in the Bermuda's and at Canary islands) were included in the simulations, covering the Atlantic Ocean part of the voyage. Additionally, simulations without and with a lower (4 m) wave height criterion were carried out as comparison. The ensemble analysis of the multiple Monte Carlo simulations showed the following results: Pi ^ÓHB ijSXtpt Fig. II: Glomar Adriatic 2 on Blue Marlin Table 6: Voyage simulation results Glomar Adriatic IV on Blue Marlin Quantity Simulations all) Di mension VMC 4.0 Operational Sign. wave height criterion m 6.0 Max. Significant Wave height 5.91 m 7.47 6.33 Max. Roll angle 16.4 11.5 degr 12.2 Max. Vert. Acceleration at bow 2.37 m/s' 3.13 2.26 Max. Transv. Acceleration CG 2.81 2.73 m/s' 3.65 Travel time range (All) 380-472 436-596 Hrs 391 436-554 Travel time (98%) Hrs 380-433 4.4 E-3 Economie (damage) risk 6.2 E-3 0.4 E-3 Human fatalities risk 0.9 E-3 2.7 E-3 Environmental risk 3.8 E-3 The simulations without using the Captain's Decision Mimic (CDM) are labelled "w.o." in the header row of Table 6. The effect of weather routing with the Hs = 6 m criterion is small: the characteristic value for the significant wave height is reduced by 5% and the voyage duration tends to be slightly longer. The reason for not finding much effect is the relatively benign climatology in January and February on the selected Southern route. A significant wave height of 6 m is seldom exceeded and if so, it is not always well avoided because the voyage is eastbound. So the prevailing weather follows the ship, which reduces the tendency to re-route or apply heading and speed changes for 'comfort'. Such observations have also been made by Van Sluijs and Stijnman in [6], The simulation result for a 4 m Hs criterion shows a clear effect: reduced risk and longer voyage duration. However, the characteristic value for the significant wave height is reduced by only 10 % for the same reason as given above. The total transverse sea-fastening force can be computed using the input option of a linear combi nat ion of signals in Safetrans. This allows the inclusion of direct (I hr mean) wind loads, while effect of wind heel is already included in the transverse acceleration ay. Hence: p (total) = M (Jack-U P ) In a similar way the leg bending moment in the jacking house can be computed, e.g.: Mx(fcti = M<«s>.h. ay<"s «'™'> _ i<"ö ^ + + p (wind on leg) L _ w (wind on leg) in which h is the distance from jacking house to the CG of (CGJack-Up) + p (wind) Applying the safety factors on the results gives the following design values for the total transverse load: Table 7: Design values using present design equation and load factors Quantity Probability of Original design Dimension value (*to yield) failure Total transverse load kN 1.0 E-2 •37490 Total transverse load kN 1.6 E-3 Transverse acceleration m/s' 4.1 The design values are given for two probabilities of failure. The advised values for consequence class C2 were not yet availabte at the submission of this paper, but are expected to be in the given range. Simulations (w.o.) 6.57 12.0 2.26 2.76 376-427 376-422 6.4 E-3 0.9 E-3 3.9 E-3 New design value (*to yield) * 29860 •41900 2.8 the leg. Inclusion of wind leads to about 6% increase ofMx<V At the issue of the paper the load factors for internal structual loads were not yet available. The actual voyage was to Alexandria from Feb. 3, 2003 to Mar. 1, 2003, of which the North Altlantic crossing was simulated as most critical part. The encountered weather and simulated route are given in Figures 12a and 12b, showing benign weather on route. •<•••,,••••,, p ^ T ^ ^ T T ^ ^ — — i ^ l ^ T B /Adrlatlc_IV_Marin/AdrlatlG_IV_on_Blu«_Mullnf 8«f«trans 31 AD Figure 12a and 12b: Voyage simulation over North Atlantic part of the voyage. 5. CONCLUSION The work carried out in the Load Factor study and Engineering Guideline for use of Safetrans is presented in this paper. Application in the example case, as well as earlier [1] and recent [7] validation work has demonstrated the applicability of the risk based, probabilistic design method. A realistic effect of weather routing is computed and can be accounted for in the design. Target applications are for heavy cargo self-propelled transports or tows, and transports of vulnerable cargo where Hmiting sea conditions appty. The linear combination of stgnals allows Safetrans to compute loads in cargo, corner loads and sea-fastening loads. The combination takes phasing of accelerations as well as wind effects properly into account. [4] Argoss B.V.: "Checking for the Dependence between Safetrans Monte Carlo Simulations" Report No. A 403, June 2004 (restricted to JIP members) [5] COMREL: Part of STRUREL for structural reliability analysis developed by RCP GmbH, Barer Strasse 48/iII, 80799 MÜNCHEN, Federal Republic of Germany. [6] M.F. van Sluijs and J.J Stijnman: "Observations on waves and ship's behaviour made on board of Dutch Ships" Netherlands Ship Research Centre TNO, Report 136 S, Dec 1971 [7] R.V. Ahilan, R. Nataraja, A.B. Aalbers, S. Anink: "SAFETRANS-Response Based Heavy Cargo Transportation Design", City University JackUp Conference, Sept. 2005 8. 6. The Authors would like to thank the following persons for their effort to make the Safetrans software a design tooi: Dr R.V. Ahilan of Noble Denton Europe for his contribution in the development of the Load Factor Study, Mr C.E.J. Leenaars of Dockwise for the contributions to the Engineering Guideline, Mr M. Levadou for his contributions to the design case calculations and Mr F. Vollen and Prof. T. Moan for their contributions to the consequence class analysis. 7. AUTHORS BIOGRAPHIES ACKNOWLEDGEMENTS REFERENCES [1] A.B. Aalbers, C.K. Cooper, S. Nowak, J.R. Lloyd, C.E.J. Leenaars and F. Vollen: " SafeTrans: A New Software System For Safer Rig Moves", City University Jack-Up Conference, Sept 2001 [2] Noble Denton Europe Ltd. "Pilot Calibration of Reliability Based Safety Factors Using V.A.C." Report No. L18753/NDE/BLC (restricted to JIP members) [3] Ocean Weather Inc.: "Wind and Wave Analysis of the IMDSS Data in Safetrans" Report to Safetrans User Group, Dec 2003 Dr R. Nataraja has over 29 years experience in Offshore Engineering. After obtaining his doctorate from Loughborough University in 1974, he was a lecturer at Cranfield Inst. of Techn., Head of R&D at Lloyd's Register of Shipping, R&D manager at Brown and Root, and at Kvaerner Earl and Wright. He has been with NDE for the last 4 years as a Sr. Principal Engineer, responsible for technology development and research projects, and carrying out Design and Concept evaluations for Certification. S. Anink holds the current position of R&D Engineer at Dockwise Transport B.V. After MSc graduation in 2001 from Delft Technical University, he is responsible for the engineering and design of non-standard transports, which include novel developments for seafastening, cribbing and design methods. A.B. Aalbers has over 25 years of experience in Offshore hydrodynamics and presently holds the position of Sr. Researcher and Joint Industry Projects Co-ordinator at the Maritime Research lnstitute Netherlands. As project manager he is responsible for the design and development of the Safetrans software and the Safetrans User Group.