Hull Stress Monitoring and Load Prediction Systems

Size: px
Start display at page:

Download "Hull Stress Monitoring and Load Prediction Systems"

Transcription

1 G. Sagvolden and K. Pran Light Structures AS, Nils Hansens vei 2, NO-0667 OSLO, NORWAY ABSTRACT Structural health monitoring (SHM) systems are recognized as a key element in a modern and efficient approach to structural lifecycle management. Fiber Optic SHM systems have been in operation in Naval vessels for 20 years, focusing on real-time measurements of global and local loads, deformations, fatigue development and early-warning systems limiting the risk of overloading events. First developed for the Royal Norwegian Navy, systems have now been installed on the platforms of several Navies and Coast guards, in addition to civilian applications in commercial shipping, oil exploration and elsewhere. In this paper, we present an overview after 20 years of experience with the technology, using examples from global load and deformation measurements, fatigue monitoring, overload predictions and early-warning systems, as well as highly dynamic ice load monitoring. 1.0 INTRODUCTION For efficient maintenance and operation of military platforms, structural health monitoring (SHM) systems are recognized as a key element in the structural lifecycle management. The potential of the fiber optic Bragg grating [1] as a sensor [2] for use in harsh environments was recognized early on, and in the mid 1990 s the Norwegian Defense Research Establishment (FFI) cooperated with the US Naval Research Labs to instrument the Royal Norwegian Navy s (RNoN) mine countermeasure vessel KNM Hinnøy [3]. This work was continued during extensive sea trials on the Royal Norwegian Navy corvette KNM Skjold [4] [5] from 1999, where a method for measuring the global loads from a network of fiber optic sensors was first applied [6]. The technology was made available by Light Structures AS, and soon adopted by commercial shipping. The first installations were on several oil tankers owned by the Navion, Mosvold and VELA shipping companies, and approved by Det Norske Veritas (DNV) for HMON class notation. In the mid 2000 s, the systems gained wider application as it was installed on the Royal Norwegian Navy Skjold series of vessels, and a large number of Liquified Natural Gas (LNG) carries with DNV HMON class or similar notation from other classification societies. Research continued on several platforms, one being the Royal Norwegian Navy s mine counter-measure vessels (MCMVs), leading to a load prediction system that is in use today on this fleet [7]. Other areas of research were the measurement of cryogenic liquid sloshing pressures in LNG tanks [8], ice load monitoring on the Norwegian coast guard vessel KV Svalbard [9] [10], dynamic responses of container vessels [11], structural textiles [12], wind turbines and aircraft structural composites [13]. Today, the systems are in active use on commercial vessels, oil and gas installations and fleets of coast guard and navy vessels. The systems are providing insight into the structural responses of these vessels to their environment and operational decisions, at different stages of the vessel lifecycle. Sometimes, sensing systems are installed during scale model tests [14] or for full scale sea trials for verification of the hull structure design STO-MP-AVT

2 basis. During operation, the systems can provide an objective measurement of the current loading conditions and relate them to the operational restrictions in terms of allowable load limits. Thus, the operators will have an objective basis for deciding when action is required, and judge the effectiveness of their actions. Further operator guidance is available through the statistical load prediction module and load predictions based on hydrodynamic vessel response models or though the database of historic loads. This database is also valuable in a condition based maintenance scheme, where inspections can be targeted and scheduled based on which loads the vessel has experienced. In this paper we will review applications and experience after 20 years of fiber optic SHM systems applied on more than 200 commercial, coast guard and navy vessels. 2.0 GLOBAL LOAD AND FATIGUE MEASUREMENTS 2.1 Global load HSM systems in commercial shipping Most HSM systems include sensors measuring the global moments. In standard design rules the limiting values for the vertical bending moment (VBM), horizontal bending moment (HBM) and twisting moment are given. The stresses from these moments are at their maximum amidships for the lowest order vibration mode. The second order mode has its highest values at the vessel quarter lengths. Thus, the basic hull stress monitoring (HSM) instrumentation recognized by all major classification societies include two sensors at main deck level amidships, and an additional sensor at each quarter length, figure 1. Figure 1: Typical minimal HSM system. This basic configuration will, naturally, only capture the most important loads. Experience shows that even this basic sensor set will give valuable insight into the vessel loads, not attainable by other means. The systems provide information on the still water vertical bending moment, thermally induced strains, wave induced loads, higher frequency dynamic responses, such as whipping and springing, and material fatigue development. In many cases, the loading instrument does not provide an accurate representation of the still water bending moment. This may be due to significant uncertainties in the mass distribution of the cargo taken onboard, such 10-2 STO-MP-AVT-305

3 as in the case of container vessels and bulk carriers. The HSM system will provide an independent measurement of the static strain values, see figure 2, and can issue a warning if the hull beam is at risk of being overloaded. Figure 2: Comparison between midships sensors (RAW, blue) and loading computer (LC, red) on a container vessel. 30-minute average values shown. Please note significant differences between the measured and loading computer values at day 20 and 160. The variations on the measured curve is due to effects not considered in the loading computer model, such as thermal gradients. Thermal gradients can also give rise to significant stresses on the structure, such as when the vessel is carrying a heated cargo, or the deck is exposed to significant sun heating. This response is captured by the HSM systems and can be analyzed for overloading and fatigue. Sea loads give rise to complex and unpredictable loading patterns at the wave encounter periods, but also at higher frequencies. For high speed vessels and vessels with large bow flare, we have seen that a significant portion of the structural response is at higher frequencies [11]. Wave impacts lead to dynamic stresses from whipping that may exceed the rule stresses both in sagging and hogging [11]. The measured high-frequency stress time series are combined with SN curves and stress concentration factors to estimate the rate of fatigue accumulation in regions close to the sensors [15]. Whipping and springing are not usually considered in fatigue life design, but SHM measurements show that these dynamic effects may contribute to more than half of the total fatigue accumulation for some vessel types [11]. The prevailing Miner-Palmgren fatigue accumulation model is non-linear in load amplitude and linear in load cycle count [15]. Thus, significant gains in vessel fatigue life may be made by minor corrections to vessel speed or course in adverse weather, reducing the load amplitudes and/or the amount of whipping/springing responses. Typically, high fatigue accumulation rates are observed during relatively short periods of time, so these corrections will have a low impact on overall transit speeds. Using the SHM system, the vessel operator receives objective and accurate information on the current rates of fatigue accumulation. The operator may set warning limits on rates of fatigue accumulation, judge the STO-MP-AVT

4 effectiveness of corrective actions, and receive information on the amount of fatigue accumulation due to whipping and springing effects. SHM systems log the load and fatigue history of the vessels which gives important information when planning inspections and maintenance. Typically, classification societies may plan the detailed inspection of hard-toaccess compartments based on the fatigue damage and overloading history of sensors in the area, potentially reducing inspection cost and time off-hire. Also, HSM data has been used for optimizing the retrofit of additional steel for strengthening the hull girder, saving time in dry dock and cost. Optimization of a SHM approach in the life-cycle management of naval ship structures has also been studied [16]. 2.2 Instrumentation HSM systems may be stand-alone, but are usually interfaced to several of the following systems: - GPS for position, speed and course - Gyro compass - Loading instruments for still-water loading condition - Wind gauge - Speed log - Centralized alarm system and voyage data recorder - Motion reference unit for vessel 6-DOF motions - Propulsion (torque, thrust) - Rudder - Mooring tension systems - Wave radars - Wave altimeters (for air-gap measurements) These connections are made to collect as much information on the environment that give rise to the monitored hull responses as possible. Detailed knowledge of the operational environment opens several possibilities for the analysis of the vessel hull responses, including benchmarking and predictions. Fiber optic hull stress monitoring technology is chosen in many SHM projects. We believe that this is due to the many advantages of fiber optic sensors, such as EMC/EMI immunity, resistance to water ingress and multiplexing several sensors on a single cable. Traditionally, HSM systems have been based on resistive strain gauges or displacement sensors, and these technologies are still used in several projects [17]. Both fiber optic and resistive strain gauge technologies require wiring, and efforts have been made to develop and demonstrate wireless alternatives [18]. 2.3 Global load measurements Although the four-sensor system shown in Figure 1 captures the major loading effects on a simple hull girder, is it not able to clearly separate the moments on the hull girder causing the hull measured hull stresses. Twenty years ago, methods were established to directly separate the moments acting on the hull girder of the RNoN corvette Skjold [6] and later also developed for the RNoN Oksøy/Alta class MCMVs [7] and model tests in a wave basin [14] STO-MP-AVT-305

5 This method relies on several sensors being placed in the mid-section of the vessel, typically using a sensor count of more than twice the number of moments to be estimated. The sensors are placed with the assistance of a finite element model (FEM) to select sensor positions and measurement directions that yield a high strain response to each load of interest, distribution of the sensors to all corners of the cross section securing maximum sensitivity to hull deflections, while avoiding positions that have a high influence of local loads. Figure 3: Fiber optic sensor positions at a cross section amidships. It is assumed that the response of the array of strain sensors, ε, is related to the moments and forces acting on the vessel by a stiffness matrix k through a set of linear equations ε = k f (1) The stiffness matrix is established by subjecting a FE model of the vessel to unit loads. Under these assumptions, the loads f can be estimated using a Singular Value Decomposition (SVD) to find the pseudo inverse of the stiffness matrix f = (k -1 ) ε (2) The strain vector ε consists of contributions from global forces, local forces and measurement noise. ε MEA = ε GLOBAL + ε LOCAL + ε NOISE (3) One advantage of the over-determined system is that excess sensors will contribute to dampen local and nose effects, in addition to providing some level of redundancy in case sensors are off line. Providing results in terms of global bending moments instead of stresses or strains at local hot-spots provides a simple and direct comparison to the design loads which both the operators and ship designers easily can relate to. The module was first used in the sea trials and design verification of the RNoN Skjold in 1999, and used in regular operation on the series of MCMVs for several years. There are several possible choices when selecting the load cases to build the basis for the linear equations in (1). Six different approaches were compared in [19] for a large displacement vessel. The technique was developed further for the case of linear seakeeping loads [20] in a non-linear time-domain simulation, STO-MP-AVT

6 demonstrating that the basis of hull deflection modes established in the linear case also can be used to construct a representative strain time-series in the case of whipping and springing loads. Often, there are many more structurally interesting spots on a vessel than those that can be instrumented within a reasonable cost and effort. On the other hand, the full richness of the vessel responses cannot be captured by simulations and calculations alone, as the actual conditions the vessel will meet are unknown and time-domain simulations that capture the non-linear sea interaction effects are very computationally intensive. Thus, if a sensor system of sufficient complexity to capture and separate the modes of the linear equations in (1) is installed, a re-combination of these modes may be used to estimate a stress time series in virtually any number of points in the vessel and analyze this time series for fatigue and overloading at several virtual sensor points. 3.0 LOCAL LOAD MEASUREMENTS SHM systems have also been used to measure structural responses to local loads. Light Structures entered into projects with DNV and other partners to develop systems measuring ice load responses [10] [21] and sloshing in LNG carrier membrane tanks [8]. The first ice load monitoring system was installed on the Norwegian Coast Guard vessel KV Svalbard in 2006 and was operated for several winter seasons in the Arctic. The measurement principle is based on measuring the shear difference of the frames supporting the shell plating in the ice interaction belt. Most instrumented frames were located in the bow, with the addition of one frame in the midships area, see figures 4 and 5. Figure 4: Instrumented frame locations on the KV Svalbard Several ice load interaction cases were modelled, confirming that the shear difference is proportional to an ensemble average of the ice loading force. γ = k i F (4) The proportionality constant, k i, depend on material parameters and geometric details of the structure. Taking the structural load capacity and safety margins into account, the shear difference can be expressed as a structural utilization which is easily understandable by the vessel operators when scaled to a percentage of the operational limit. A 24 hour structural utilization history for the L1 bow frame is shown in figure STO-MP-AVT-305

7 Figure 5: Instrumented frame locations on the KV Svalbard Figure 6: History of the structural utilization of the L1 bow location when operating partly in ice and open waters A similar system was installed on the South African polar supply and research vessel S.A. Aghulas II in 2012 [22] and has resulted in an active research project spanning several years [23]. Similar systems have been installed on the new Norwegian polar research vessel Kronprins Haakon, an LNG tanker operating in the Arctic, and additional vessels currently under construction. The shear difference approach to ice load measurements can be formulated as a strain to load conversion, and solved using the same principles as (1). These methods have also been applied to more complex structures, such as the response of an LNG containment system to liquid sloshing inside the cargo tanks and the forces acting on the columns and pontoons of a semisubmersible offshore platform. The semisubmersible was instrumented on 80 locations as shown in Figure 7. Stiffness matrices were established for loads acting on each column and pontoon, combined with global forces acting on the entire platform submersed structure. STO-MP-AVT

8 Figure 7: The sensor network installed on the columns and pontoons of a semisubmersible offshore platform. The illustration shows one of four columns, and half of one pontoon. 4.0 LOAD PREDICTIONS A method predicting extremal loads in near future was developed for the RNoN fleet of MCMVs [7]. In this method, the peak-over-threshold observations of the vertical bending moment are fitted to an extremal value distribution. After analysis of data from the pilot vessel in the series, the Generalized Pareto Distribution was selected as the best suited statistical model. This model allows estimation of the most likely maximum load (return value) within a certain observation period (return period), including the confidence intervals of the estimate. The threshold used for identifying the extreme loads is continuously updated based on the current observations. The sea conditions are also monitored, resetting the peak load register whenever a significant change in the operational environment is detected to ensure that the load predictions are based on representative observations. An illustration of the user interface is shown in Figure 8. The load history for the latest 20 minutes is shown as a green curve, while the current most likely maximum load (return value) for the return period is shown in red. An alarm is triggered when the predicted load crosses the allowable limits (orange lines), giving the vessel operator an opportunity to change course and speed to potentially avoid extreme loads. In this example, no action was taken and a hogging load exceeding the allowable limit was observed a few minutes after the prediction was made. This mechanism allows for setting sailing restrictions based on objective measurements directly related to loads and risk of damage, instead of a subjective assessment of the sea conditions. Extreme value statisticsbased load prediction has now been included in the latest DNVGL rules for hull stress monitoring systems, and we expect that this will contribute to the safety of vessels operating along the Norwegian coast and other similar environments. Extreme value statistics may also be used to predict maximum local loads, and promising results were obtained in the case of ice loads on the KV Svalbard [10]. Similar results have been shown for sloshing pressures in LNG tanks STO-MP-AVT-305

9 Sagging Real-time input Hogging Approx. 20min Figure 8: The vertical axis is the hull loading value. The horizontal axis is time with new data entering from right to left. The length of the time axis is adjustable (approx. 20min in this example). The green line is the measured hull load. The red lines are the statistically predicted VBM (sagging + / hogging -). The thick orange lines are the alarm trigger levels. From [7]. 5.0 SUMMARY AND CONCLUSIONS This paper has presented some of the applications and developments in structural health monitoring on Navy, Coast Guard and commercial vessels since the first trials with fiber optic hull stress monitoring systems 20 years ago. Most systems have been designed to monitor global loads, mainly the vertical bending moment, and fatigue development. Several systems include extra sensors to separate the global load components or provide detailed information on local loads. Methods for predicting future extreme loads have been in operation for several years and allow a more flexible and reliable risk-based operation of vessels supplied with this equipment. Recent developments indicate a wider use of SHM data in the near future. Efforts are currently made to gather data from several SHM equipped vessels in large databases for analysis, benchmarking and reporting. FEM and methodologies are being developed to further optimize sensor placement, and extract information on strain and fatigue from hundreds of virtual sensor locations. Navies and Coast Guards are installing SHM systems on fleets of vessels to improve maintenance planning and fleet management. STO-MP-AVT

10 REFERENCES [1] G. Meltz, W. Morey and W. H. Glenn, "Formation of Bragg gratings in optical fibers by a transverse holographic method," Optics letters, vol. 14, no. 15, pp , [2] C. Campanella, A. Cuccovillo, C. Campanella, A. Yurt and V. Passaro, "Fibre Bragg Grating Based Strain Sensors: Review of Technology and Applications," Sensors, vol. 18, no. 9, p. 3115, [3] A. Kersey, M. Davis, T. Berkoff, A. Dandridge, R. Jones, T. Tsai, G. Cogdell, G. Wang, G. Havsgaard, K. Pran and S. Knudsen, "Transient load monitoring on a composite hull ship using distributed fiber optic Bragg grating sensors," Smart Structures and Materials, vol. 3042, pp , [4] G. Wang, K. Pran, G. Sagvolden, G. B. Havsgård, A. E. Jensen, G. A. Johnson and S. T. Vohra, "Ship hull structure monitoring using fibre optic sensors," Smart materials and structures, vol. 10, no. 3, pp , [5] K. Pran, G. Johnson, A. Jensen, K. Hegstad, G. Sagvolden, O. Farsund, C. Chang, L. Malsawma and G. Wang, "Instrumentation of a high-speed surface effect ship for structural response characterization during sea trials," Smart Structures and Materials, vol. 3986, pp , [6] A. Jensen, J. Taby, K. Pran, G. Sagvolden and G. Wang, "Measurement of global loads on a full-scale SES vessel using networks of fiber optic sensors," Journal of ship research, vol. 45, no. 3, pp , [7] A. Jensen, J. Taby, H. Torkildsen, P. Brodtkorb, S. Løvseth, G. Sagvolden and K. Pran, "Safe operation of ships using real-time monitoring and statistical predictions," in RTO-AVT-164, Bonn, [8] Ø. Lund-Johansen, T. Østvold, C. Berthon and K. Pran, "Full scale measurements of sloshing in LNG tanks," in 25th Gastech Conference, Amsterdam, [9] G. Sagvolden and K. Pran, "Structural Model Systems with applications to Ice Response Monitoring," in MTS Dynamic Positioning Confence, [10] B. Leira, L. Børsheim, Ø. Espeland and J. Amdahl, "Ice-load estimation for a ship hull based on continuous response monitoring," Proceedings of the Institution of Mechanical Engineers, Part M: Journal of Engineering for the Maritime Environment, vol. 223, no. 4, pp , [11] M. Barhoumi and G. Storhaug, "Assessment of whipping and springing on a large container vessel," International Journal of Naval Architecture and Ocean Engineering, vol. 6, no. 2, pp , [12] Polytect project, "Polyfunctional Technical Textiles against Natural Hazards," cordis, [13] G. Sagvolden and T. Thorvaldsen, "Structural Health Monitoring of Scarf Repaired Military Aircraft Structures Using an FBG Based Sensor System," in STO-AVT-266, Turin, [14] A. Jensen, J. Taby, K. Pran, A. Pedersen and E. Jullumstrø, "Global load estimations for a 75 meter FRP Composite SES vessel using a scaled model instrumented with a network of fiber optic sensors," in 11th International Conference on Fast Sea Transportation, Honolulu, [15] DNV, "Fatigue assessment of ship structures," DNV, [16] D. M. Frangopol, "Integrated Life-Cycle Framework for Maintenance, Monitoring and Reliability of Naval Ship Structures," Lehigh University, [17] I. Drummen, M. Schiere, R. Dallinga and K. Stambaugh, "Full and Model Scale testing of a New Class of US Coast Guard Cutter," in Ship Structures Committee Symposium, Linthicum Heights, Maryland, [18] R. Swartz, A. Zimmerman, J. Lynch, J. Rosario, T. Brady, L. Salvino and K. Law, "Hybrid wireless hull monitoring system for naval combat vessels," Structure and Infrastructure Engineering, vol. 8, no. 7, pp , STO-MP-AVT-305

11 [19] F. Bigot, Q. Derbanne and E. Baudin, "A review of strains to internal loads conversion methods in full scale measurements," in PRADS, Changwon City, Korea, [20] F. Bigot, F. Sireta, E. Baudin, Q. Derbanne, E. Tiphine and Š. Malenica, "A Novel Solution to Compute Stress Time Series in Nonlinear Hydro-Structure Simulations," in ASME th International Conference on Ocean, Offshore and Arctic Engineering, St John's, NL, Canada, [21] G. Sagvolden and K. Pran, "Ice Response Monitoring Using Structural Monitoring Systems," in Proceedings of the International Conference on Port and Ocean Engineering Under Arctic Conditions, [22] A. Bekker, M. Suominen, O. Peltokorpi, J. Kulovesi, P. Kujala and J. Karhunen, "Full-scale measurements on a polar supply and research vessel during maneouver tests in an ice field in the Baltic Sea," in ASME rd International Conference on Ocean, Offshore and Arctic Engineering, [23] A. Bekker, M. Suominen, P. Kujala, R. De Waal and K. Soal, "From data to insight for a polar supply and research vessel," Ship Technology Research, pp. 1-17, STO-MP-AVT

12 10-12 STO-MP-AVT-305