Application of artificial neural network in fixed offshore structures

Transcription

1 Indian Journal of Geo-Marine Sciences Vol 44(3), March 2015, pp. Review article Application of artificial neural network in fixed offshore structures Md. Alhaz Uddin*, Mohammed Jameel & Hashim Abdul Razak Department of Civil Engineering, University of Malaya, Kuala Lumpur, Malaysia *[ Received 07 July 2012; revised 08 November 2012 Artificial Neural networks (ANNs) are good for prediction, damage detection, online monitoring and controlling in fixed jacket platform in such cases where formal analysis is tough or impossible. This paper presents state-of-the-art information reported so far on the research studies of ANN applications and future opportunity for fixed jacket offshore platforms. It is found that ANNs provide fairly accurate results as compared to conventional method. Based on the review, the future scope of work is outlined. [Keywords: Artificial Neural network, Fixed jacket platform, Shallow water, Damage detection, Online monitoring, Prediction, Controlling] Introduction Several types of offshore structures are in use for oil and gas exploration due to global energy demand. Fixed types (jacket and gravity) offshore structures are economically viable for shallow water regions. Period of their service life depends on various environmental forces 2. Numerous studies have been carried out by conventional methods for analysis of offshore platforms 3-9. Artificial Neural Networks 10 can calculate response in rapid mode. Neural networks are compatible to solve problems associated with fixed jacket platform. ANNs are currently being successfully used for a wide-range of problems in various fields 11 such as business, industry and science 12. It helps in pattern recognition, pattern classification, time series prediction, process control, and process modeling. Neural networks are incorporated in offshore engineering and stability analysis of the rubble mound breakwater 13, identifying the crack extent and location of damage on the longitudinal faceplate of ship s structure 14. Miri et al. 15 presented two different types of ANN, which predicts the risk of different pipe sticking in the Iranian offshore oil field. Makarynskyy 16 has discussed improvement of wave predictions with neural network. Mahfouz 17 presented a successful method of predicting the capability-polar-plots for dynamic positioning systems of offshore platforms. Many researchers have used ANN to address various types of problems in fixed platforms. Problems such as structural damage detection 18-20, damage localization 21, fatigue 22-24, response control 25, vibration control 26, prediction of force and moment 27 has been presented. The main objective of this study is to deal with a detail review on solution progression of fixed jacket platforms by implementing ANN. It has a huge potential in future, and is expected to greatly enhance the efficiency of overall production and operations activities. Artificial Neuron Model Artificial neural networks (ANNs) are an information-processing system encouraged by biological nervous systems, such as human brain. ANNs are able to learn from experience like people. An ANN is a combination of artificial neurons namely processing elements, nodes or units. Each processing element is fully interconnected to other processing elements by its connection weights. Processing element accepts its weighted inputs, which are summed with adjustable unit bias. The bias unit is utilized to scale up the input to develop the convergence properties of the neural network. The result of this combined summation is passed through the transfer function to produce the output of the processing element.

2 UDDIN et al.: APPLICATION OF ARTIFICIAL NEURAL NETWORK IN FIXED OFFSHORE STRUCTURES The artificial neurons may have single input or multiple inputs as described below. Single Input Neuron- Fig. 1 shows that a single input artificial neuron model in its simplified mathematical form which has been extracted with some modification from Shahin et al., 28. X W Fig. 1 Single input artificial neuron model I f(i) I WX (1) where, X= Input W= Connection weight θ = Bias I = Produced response f(.) = Transfer function O = Output. As fig. 1 shows the input neuron X is multiplied by connection weight W to form WX. The bias θ passed into the summing junction and produced O response I which is shown in Equation 1. The produced response I move into the transfer function f (I) and give an output O. Multiple Input Neuron- A typical multiple input neuron with a transfer (activation) function which has been extracted with some modification from Shahin et al., 28 is shown in Fig. 2. Neurons usually have more than one input which is multiplied with, each input having its own weight. The adjustable unit bias passed into the summing junction and produced response which is shown in Equation 2. The produced response transfers through the activation function and gives output. I j j n i1 W ji X i where, X i = Input from node i, i = 0, 1,, n W ij = Connection weight between nodes j and i θ j = Bias for node j I = Produced response f(.) = Transfer function O j = Output. (2) Transfer (Activation) Function: Activation function is contained in hidden layer in between inputs and outputs. This function is also called as transfer function. Finally, an activation function controls the amplitude of the output. Each transfer function has its own formula which is shown in Table 1. X 1 W j1 I j j n i1 W X ji i f ( I j ) O j X n W jn j Fig. 2 Multiple inputs artificial neuron model

3 INDIAN J MAR SCI, VOL 44(3), MARCH 2015 Table 1 Different types of transfer (activation) function Function Mathematical Representation Threshold (hard-limit) Sigmoid Hyperbolic Tangent Gaussian Binary 1 if i 0 f I 0 if i 0 Bipolar 1 if i 0 f f f f I 1 if 1, 1 i e i 0 I 0 1 e 1 e vi I tanh( vi ), v 0 I vi 2 ( i ) e Here α is the slope parameter also called shape parameter; the v is also used for same representation. Classification of ANNs: ANNs can broadly be divided into the following three categories, based on the arrangement of neurons and the connection patterns of the layers: Feed-forward neural networks, Feed-back neural networks, and Self-Organizing Maps. Feed-forward neural network A feed-forward neural network is a standard type of neural network that might be used for applications such as prediction, control and monitoring. Feed-forward neural networks are oneway connections (weights) from input to output layers through the hidden layer 29,30. Feed-forward networks often have one or more hidden layers followed by an output layer. This network propagates in forward direction, it never goes in backward direction 31. In a feed-forward neural network, the input numbers are served to input nodes, after multiplying by a weight it goes in the hidden layer nodes. A hidden layer node receives the weighted input from each input node, and it connects with bias. Then the result passes through the nonlinear transfer function. The output nodes come from the hidden layer node. The network is trained by trial and error, the target output at each output node is compared with the network output and the difference (error) is minimized by adjusting the weights and biases though training algorithms. The obtained optimum network can be used for various applications within the range of trained data 32. Multilayer Perceptron-Multilayer perceptron (MLP) is a feed-forward artificial neural network model that transforms sets of input signals into sets of output signals through hidden layer 33. The hidden layer and nodes play especially significant part for many successful applications of neural networks 34. In MLP, the input signal on a layer-bylayer basis propagates in forward direction through the network. The network is trained in supervised learning method with error back propagation algorithm. Feed-back neural network A feed-back network is a network with connections from the output of a layer to its input. The feed-back connection can be direct or can pass through several layers. In feed-back networks, the output information defines the initial activity state of a feed-back system, and after the state transitions to the asymptote, final state is identified as the outcome of the computation. The networks known as recurrent networks are feed-back networks. Self-organising maps The self-organising map describes a mapping from a high dimensional input space to a low dimensional map space and is trained using unsupervised learning. Self-organising maps are different from other artificial neural networks in the sense that they use a neighborhood function to preserve the topological properties of the input space. An ANN can also be self-organising which means that it can create its own organisation or

4 UDDIN et al.: APPLICATION OF ARTIFICIAL NEURAL NETWORK IN FIXED OFFSHORE STRUCTURES Fig. 3 Classification of learning algorithms 1 representation of the information it receives during learning time without any intervention from the outside world. Learning Algorithm: The neural network is trained to adjust the values of the connections (weights) to execute a particular function between an exact input response and a precise target output. This process is known as learning. There are three basic types of learning methods in neural network such as supervised, unsupervised and reinforced learning which is shown in Fig. 3. Among them supervised and unsupervised learning algorithm are most popular 1. In supervised learning, the target response of the network is compared to a desired output response. If the real target response differs from the desired output response, the network generates an error. This error is utilized for adjusting the connection weights between the target and output response, till it matches with actual outputs 35. Unsupervised learning does not require desired target response. The connection weights are adjusted with network according to the input values. In reinforced learning, the target response of the network is not compared to a desired output response. But only indicates that the computed output is correct or incorrect. The information provided helps the network in its learning process. Back-propagation Algorithm: Back-propagation algorithm is most popular for training the neural networks. It requires lower memory than other algorithms and commonly makes acceptable error level. This algorithm could be used in various types of networks even though it is usually suitable for training MLPs 36. It was developed by simplifying the Widrow-Hoff learning rule for multiple-layer networks and nonlinear differentiate transfer functions. Input and the corresponding target are used to train the network until it gives fairly accurate function. The results and calculated errors are passed forward from input nodes to output nodes 37,38. Performance Functions: The performance function is used to compute network performance in training. This is useful for many algorithms, such as back-propagation, which operate by adjusting network weights and biases to improve performance 39. The following functions are conventional performance functions: MAE =mean absolute error MSE =mean of squared errors SSE =sum of squared errors

5 INDIAN J MAR SCI, VOL 44(3), MARCH 2015 Table 2 Summary of offshore structural research using ANN Researchers Types of work Input variable No. of hidden layer/neuron Transfer fun. hidden: output Training Algorithm Outputs Mangal et al. 19 Online monitoring Varies on networks 3/NA hard limiters, sigmoid: hyperbolic tan BPN & ART 1 Lopes and Ebecken 24 Diao et al. 21 Elshafey et al. 18 Chang et al. 25 Liang et al. 26 Ma et al. 40 Fatigue monitoring Damage localization Damage detection Response control Vibration control Optimal control 14 2/8 Sigmoid: sigmoid FFBPN 1 3 3/NA Sigmoid: sigmoid BPN & PNN 1 2 1/NA Sigmoid: sigmoid BPN 1 2 1/ 5 NA MLPBPN 1 3 NA NA FFOC 1 1 NA NA FFOC 1 Elshafey et al. 27 Prediction of force and moment 3 1/50 hyperbolic tangent: sigmoid, linear FFN 2 NA-Not Available, BPN-Back-propagation network; ART-Adaptive resonance theory; PNN-Probabilistic neural network; FFOC-Feed-forward and feedback optimal control; MLPBPN-Multilayer perceptron back-propagation network; FFN-Feed-forward network. ANN for Offshore Structures Artificial Neural Networks (ANNs) are widely used across all disciplines of coastal, ocean, marine and offshore engineering. Mahfouz 17 presented a method for predicting the capability-polar-plots of dynamic positioning systems of offshore platforms, stability analysis of the rubble mound breakwater 13, estimating the cost, uncertainties and risk analysis of floating structures 41, and identifying the crack extent and location of damage on the longitudinal faceplate of ship s structure 14. ANN are also used to predict the risk of different pipe sticking in the Iranian offshore oil field 15. Table 2 shows the summary of research in fixed offshore platform which was carried out by various researchers during Several aspects such as online monitoring, fatigue monitoring, damage localization, damage detection, response control etc. has been studied. ANNs are used for these types of problems where formal analysis is time consuming and complex. The table shows number of input variables, number of hidden layers and type of transfer function used. It helps to select the number of hidden layers and suitable type of activation functions for various aspects of fixed platform. Fixed Offshore Structure Fixed jackets are the most popular offshore structures for shallow-water applications. Among the various types of existing platforms all over the world, approximately 95% installed platforms are fixed jacket 42,43. Because these structures use proven technology, their machineries are easily available over the range providing reliable and cost effectiveness. The fixed platform is economical for

6 UDDIN et al.: APPLICATION OF ARTIFICIAL NEURAL NETWORK IN FIXED OFFSHORE STRUCTURES installation in depth of water up to 400 m 44. With increase in water depth, fixed offshore structures became uneconomical and unstable. Damage Detection for Fixed Jacket: Offshore structures are prone to damage due to various types of environmental forces which exceed the design load during operation. All kinds of damage should be identified quickly so that corrective action can be taken to overcome disastrous failures. For that reason, it is expected to ensure reliable monitoring and competent system 45 for such structures. Wu et al., developed automatic monitoring methods using neural network for detection of structural damage. Backpropagation neural network was used to recognize the behavior of the undamaged structure and also observed the behavior of the structure with different probable damage states. It was found that the developed monitoring method is able to detect any existing damage. Mangal et al., 19 proposed adaptive resonance theory (ART) for online monitoring to detect damage and location of fixed jacket platform. BPN and ART network were used for comparison of damage detection. ART networks work as pattern recognition, has the ability to be train new data or adapt themselves to new conditions. Speech recognition and object identification problem can be explained easily with BPN. It was suitable to use both BPNs and ART simultaneously for good results. Banerji and Datta, 46 used ANN for monitoring the integrity of offshore structures. Feed forward neural network algorithm was used for pattern recognition and comparing measured RMS displacement responses at different elevations of an offshore structure for various sea states. Lopes and Ebecken 24 proposed a method using neural network for fatigue monitoring of fixed offshore structures so that an automatic fatigue data gaining on board and performing in-time fatigue calculation to the actual loading condition can be achieved. Finite element method and stochastic fatigue analysis were used for fatigue damage calculation of the 8 brace locations of each selected joints during 20-minute periods. The results clearly indicated that the fully connected network produces small errors and better performance. A laboratory jacket platform model to access the feasibility of adapting vibration responses due to impulse and relaxation for on-line structural monitoring has been investigated by Mangal et al., 47. Effects of damage in six members of the platform and changes in deck masses were studied. A finiteelement model of the structure was used to analyze all the cases for comparison of results and system identification. Zhen and Zhigao 20 proposed the time-domain response data with measurement noise under random loading for detecting damage of offshore platforms. A sensitivity matrix consisting of the first differential of autoregressive coefficients of timeseries models with respect to the stiffness of the structural elements was obtained based on time domain response data. Numerical simulations show that the use of few sensors acceleration history data with certain level measurement noises is capable of detecting damages efficiently, and the increase in numbers of sensors helps in improving the diagnosis success rate. The damage localization of the fixed platform has been presented utilizing probabilistic neural networks (PNN) by Diao et al. 21. The members of offshore platforms were classified and separated into several layers and the decision system for the type and layer of damaged members was established using the back-propagation network (BPN). The experimental results show that the trained neural networks are able to detect the damages with reasonable accuracy. Damage detection under random loads were evaluated by the combined method of random decrement signature and neural networks by Elshafey et al. 18. The random decrement technique was used to extract the free decay of the structure from its online response while the structure was in service. The free decay and its time derivative were used as input for a neural network. The output of the neural network was used as an index for damage detection. It has been shown that function N (number of segments) was effective in damage detection in the members of an offshore structure. Controls for Fixed Jacket: The safety of offshore structures is a significant concern under the oceanic environmental loads. The reduction of the dynamic response of offshore structures subjected to random ocean waves has been a critical issue in terms of serviceability, fatigue life, and safety of the structure. A number of studies have been carried out to control vibrations in offshore structures using ANN. Modified probabilistic neural network (MPNN)

7 INDIAN J MAR SCI, VOL 44(3), MARCH 2015 was used for controlled response of fixed offshore structures subjected to random ocean waves 25. Linear quadratic regulator (LQR) algorithm was used for calculation of control forces under random ocean wave. Significant decrease in structural responses proves the MPNN control algorithm as an effective vibration control technique. Nonlinear interactions of pile-soil by wave load disturbances of offshore jacket platforms was studied by Liang et al. 26. Morison equation was used to estimate the wave load disturbances. Results of feed-forward and feedback vibration control jackettype offshore platform were significantly effective. The vibration control was studied with an active mass damper 48 and presented a feed-forward and feedback optimal control (FFOC) under irregular waves 40. The reductions of lateral deck motion under different control laws were compared. The simulation shows that FFOC was more efficient and robust than the classical feedback optimal control (CFOC). Hong-yu and De-you 49 proposed grey neural network as adaptive predictive inverse controller, which was implemented to the active control of jacket offshore platform. The simulation results show that grey neural network has strong robustness, and can effectively control the displacement of jacket offshore platform under random loads. Ya-jun et al. 50 presented classical optimal control using artificial neural networks for structural control of offshore platforms under random waves. BPN algorithm was used for active control. Result shows that the active control was feasible and effective. Cui and Zhao 51 proposed an approach of rough neural network as an adaptive predictive inverse controller for active control of offshore platform under the combined action of random waves and winds. The result showed that the proposed rough neural networks were fast in training speed and strong in robustness. The displacement response of platform top was effectively controlled and the time delay was sufficiently suppressed. This method also possesses strong anti-disturbance capability. An adaptive inverse control method on the basis of novel rough neural networks (RNN) to control the harmful vibration of offshore jacket platform was proposed by Hong-yu et al. 52. The offshore jacket platform model was established by dynamic stiffness matrix (DSM) method. The constructed novel RNN has advantages such as clear structure, fast training speed and strong error-tolerance ability. Furthermore, the proposed method based on RNN could effectively control the harmful vibration of the offshore jacket platform. A new neuro-control scheme was applied to the vibration control of a fixed offshore platform under random wave loads to examine the applicability of the Lattice Probabilistic Neural Network (LPNN) 53. The results of LPNN showed better performance in effectively suppressing the structural responses in a shorter 54 computational time. Wei & Gong-you investigated the optimal control for linear systems affected by external harmonic disturbance and applied to vibration control systems of offshore steel jacket platforms using ANN. It was established that the control scheme was useful in reducing the displacement response of jacket-type offshore platforms. The control method for fixed offshore platforms using semi-active tuned liquid column damper (TLCD) was presented by Hong-nan and Lin-sheng 55. A back propagation artificial neural network was used to adjust the orifice opening of TLCD because of the nonlinear motion of liquid in TLCD. Ya-jun and De-you 56 proposed a new active control scheme on the basis of neural network for the suppression of oscillation in multiple-degree-offreedom (MDOF) offshore platforms. The obtained results show that the active control was feasible and effective, and it finally overcomes time delay owing to the robustness, fault tolerance, and generalized capability of artificial neural network. Response Prediction of Fixed Jacket Platform using ANN: Offshore structures are subjected to external environmental load such as wind loads, wind generated wave excitations, and current forces. The dynamic response prediction of offshore structures in a random environment is a significant aspect of design 57. ANN has been successfully used in offshore structures for prediction of dynamic characteristics 58 by limited number of researchers. Elshafey et al. 27 provided a deck acceleration measurement tool to predict the value of the force and moment acting on the offshore structure foundation subjected to wind generated wave excitation. Neural networks and Fokker-Planck formulation methods were used to determine the

8 UDDIN et al.: APPLICATION OF ARTIFICIAL NEURAL NETWORK IN FIXED OFFSHORE STRUCTURES relationship between the force and moment acting on the foundation and deck acceleration. The total virtual mass of the equivalent single degree of freedom formulation of the structure was determined at different deck masses. Yun and Bahng 59 used back propagation neural networks (BPNN) to estimate the number of unknown parameters of jacket platform. The jacket-type offshore structure was presented to illustrate the proposed procedure and to demonstrate the effectiveness of the method. Mandal et al. 60 addressed the prediction of stress resultant deflection of fixed offshore platform under varying sea environments using neural networks. With the data of deck displacements at various loading states, the Back-propagation neural network was first trained and later prediction of the deck displacements is obtained for any loading state. The result showed that prediction of deck displacement was significantly accurate. Yasseri et al. 61 proposed a predictive method for identifying the range of sea-states considered safe for the installation of offshore structures using FEM method and ANN. A finite element dynamic analysis was performed and a table of some safe and unsafe sea states has been prepared based on the pile allowable stress. Multi-layer feed-forward neural networks were used to determine whether the predicted sea state is safe or not. Conclusion This paper gives a detail literature review on the fixed offshore structures using ANN. It includes the research on damage detection, online monitoring, control, and response prediction of shallow water structure. In addition, this study attempted to address the details of ANN configuration. ANNs have satisfactory advantages over conventional computing techniques. However, there are no structured methods today to identify the network structure which is best to approximate the function and mapping the inputs to outputs. Hence, the tedious experiments and trial-and-error procedures are often used. ANNs can deal with more data and requires less computational time for training. The design parameter prediction such as shear force, torsion and deflection for fixed jacket structural elements by ANN is still young. Furthermore, long term decay by sea water and corrosion experienced in such structures are to be explored in optimum manner. ANNs can ensure itself as competent tool for such predictions. Acknowledgement Authors gratefully acknowledge University of Malaya (UM), for supporting this research through Grant no. RP004E-13AET. Reference 1 Chakraborty, R.C., Fundamental of Neural Networks:Soft Computing Course Lecture 7-14, notes, slides, Available from Internet:< (2010): 2 Kumar, B.P., Reliability based design method for coastal structures in shallow seas, Indian J. of Geo-Marine Sci., 39 (2010) Chaudhury, G.K. & Dover, W.D., Fatigue analysis of offshore platforms subject to sea wave loadings, Inter. J. of Fatigue, 7 (1985) Elshafey, A.A., Haddara, M.R. & Marzouk, H., Dynamic response of offshore jacket structures under random loads, Marine Struc., 22 (2009) Kareem, A., Hsieh, C. & Tognarelli, M., Frequencydomain analysis of offshore platform in non-gaussian seas, J. of Eng.Mech, 124 (1998) Li, H., Wang, J. & James Hu, S.-L., Using incomplete modal data for damage detection in offshore jacket structures, Ocean Eng., 35 (2008) Park, M., Koo, W. & Kawano, K., Dynamic response analysis of an offshore platform due to seismic motions, Eng. Struc., 33 (2011) Sawaragi, T. & Nochino, M., Two dimensional responses of fixed offshore platform in waves, Applied Ocean Research, 7 (1985) Geng, B.-L., Teng, B. & Ning, D.-Z., A time-domain analysis of wave force on small-scale cylinders of offshore structures, J. of Marine Sci. and Technology, 18 (2010) Garstecki, A., Knitter Piatkowska, A., Pozorski, Z. & Ziopaja, K., Damage detection using parameter dependent dynamic experiments and wavelet transformation, J. of Civil Eng. and Manag., 10 (2004) Deo, M.C., Artificial neural network in coastal and ocean engineering, Indian J. of Geo-Marine Sci., 39 (2010) Widrow, B., Rumelhart, D.E. & Lehr, M.A., Neural networks: applications in industry, business and science, Communications of the ACM, 37 (1994) Mandal, S., Rao, S., Manjunath, Y. & Kim, D., Stability analysis of rubblemound breakwater using ANN, in Fourth Indian National Conf. on Harbour and Ocean Eng. (INCHOE-2007), National Institute of Technology, Surathkal, Surathkal, India, (2007): Zubaydi, A., Haddara, M.R. & Swamidas, A.S.J., Damage identification in a ship s structure using neural networks, Ocean Eng., 29 (2002)

9 INDIAN J MAR SCI, VOL 44(3), MARCH Miri, R., Sampaio, J., Afshar, M. & Lourenco, A., Development of Artificial Neural Networks to Predict Differential Pipe Sticking in Iranian Offshore Oil Fields, in Intern. Oil Conf. and Exhibition, Mexico, (2007): 16 Makarynskyy, O., Makarynska, D., Kuhn, M. & Featherstone, W.E., Predicting sea level variations with artificial neural networks at Hillarys Boat Harbour, Western Australia, Estuarine, Coastal and Shelf Sci., 61 (2004) Mahfouz, A.B., Predicting the capability-polar-plots for dynamic positioning systems for offshore platforms using artificial neural networks, Ocean Eng., 34 (2007) Elshafey, A.A., Haddara, M.R. & Marzouk, H., Damage detection in offshore structures using neural networks, Marine Struc., 23 (2010) Mangal, L., Idichandy, V.G. & Ganapathy, C., ARTbased multiple neural networks for monitoring offshore platforms, Applied Ocean Research, 18 (1996) Zhen, W. & Zhigao, Z., Damage detection of offshore platform structures using time domain response data, in Intern. Conf. on Intelligent Computation Tech. and Autom., IEEE, Changsha, Hunan, China (2010): Diao, Y.-S., Li, H.-J., Shi, X. & Wang, S.-Q., Damage localization for offshore platform by neural networks, in Machine Learning and Cybernetics, Proceedings of 2005 Intern. Conf., 8 (2005): Lopes, P., Fatigue damage prediction via neural networks, Offshore Eng., Computational Mechanics Publications, Boston(USA), (1997) Lopes, T.A.P., Neural networks on fatigue damage prediction, in Intern. Conf. on Neural Networks,1997.,, IEEE, Houston, TX, USA 1(1997): vol Lopes, T.A.P. & Ebecken, N.F.F., In-time fatigue monitoring using neural networks, Marine Struc., 10 (1997) Chang, S., Kim, D., Chang, C. & Cho, S.G., Active response control of an offshore structure under wave loads using a modified probabilistic neural network J. of Marine Science and Tech., 14 (2009) Liang, Y.J., Tang, G.Y. & Zhang, C.M., Feedforward and feedback control for nonlinear systems of offshore platforms, IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, (2010) Elshafey, A.A., Haddara, M.R. & Marzouk, H., Estimation of excitation and reaction forces for offshore structures by neural networks, Ocean Systems Eng., 1 (2011) Shahin, M.A., Jaksa, M.B. & Maier, H.R., State of the art of artificial neural networks in geotechnical engineering, Electronic J. of Geotech. Eng., 8 (2008) Mohammadhassani. M, Jumaat M.Z, Nezamabadi-pour H, Jameel Mohammed & Arumugam, A.M.S., Application of artificial neural network (ANN) and linear regressions (LR) in predicting the deflection of concrete deep beams, Computer and concrete, 11 (2013) Beale, M.H., Hagan, M.T. & Demuth, H.B., Neural Network Toolbox 7 User s Guide., (2010): 31 Khanchandani, K.B. & Hussain, M.A., Emotion recognition using multilayer perceptron and generalized feed forward neural network, J. of Sci. & Industrial Research, 68 (2009) Rai, R. & Mathur, B., Event-based Sediment Yield Modeling using Artificial Neural Network, Water Resources Manag., 22 (2008) Manik, A., Gopalakrishnan, K., Singh, A. & Yan, S., Neural networks surrogate models for simulating payment risk in pavement construction, J. of Civil Eng. and Manag., 14 (2008) Zhang, G. & Eddy Patuwo, B., Forecasting with artificial neural networks: The state of the art, Intern. J of forecasting, 14 (1998) Lefik, M. & Schrefler, B., Artificial neural network as an incremental non-linear constitutive model for a finite element code, Computer Methods in Applied Mechanics and Eng., 192 (2003) Dikmen, S.U. & Sonmez, M., An Artificial Neural Networks Model for the Estimation of Formwork Labour, J. of Civil Eng. and Management, 17 (2011) Tripathy, R.R. & Maity, D., Damage assessment of structures from changes in curvature damage factor using artificial neural network, Indian J. of Eng. & Materials Sciences, 11 (2004) Rajesh, K. & Ray, A.K., Artificial neural network for solving paper industry problems: A review, J. of Scientific & Industrial Research, 65 (2006) Makarynskyy, O., Artificial neural network in merging wind wave forecasts with field observations, Indian J. of Marine Sci., 36 (2007) Ma, H., Tang, G.-Y. & Zhao, Y.-D., Feedforward and feedback optimal control for offshore structures subjected to irregular wave forces, Ocean Eng., 33 (2006) Cocodia, E.O., Creating Better Cost Estimates for Floating Offshore Structures by Assessing Cost Correlation and Understanding Risk, in The Eighteenth (2008) Intern. Offshore and Polar Eng. Conf., International Society of Offshore and Polar Engineers, P. O. Box 189, Cupertino, CA, , USA, Vancouver, BC, Canada, (2008): 42 Wang, Y., Duan, M., Wang, D., Liu, J. & Dong, Y., A Model for Deepwater Floating Platforms Selection Based on BP Artificial Neural Networks, in International Offshore and Polar Eng. Conf., Maui, Hawaii, USA, (2011): Twomey, B.G., Study assesses Asia-Pacific offshore decommissioning costs, in Oil and Gas J. (2010): DNV, Guideline for Offshore Structural Reliability: Application to Jacket Platform, in Joint Industry Project, Det Norske Veritas Norway, (1996): Wu, X., Ghaboussi, J. & Garrett, J.H., Use of neural networks in detection of structural damage, Computers & Struc., 42 (1992) Banerji, P. & Datta, T.K., Integrity Monitoring of Offshore Platforms Using Artificial Neural Nets, in Proceedings of the Seventh International Offshore and Polar Eng. Conf., Honolulu, USA, IV (1997):

10 UDDIN et al.: APPLICATION OF ARTIFICIAL NEURAL NETWORK IN FIXED OFFSHORE STRUCTURES 47 Mangal, L., Idichandy, V.G. & Ganapathy, C., Structural monitoring of offshore platforms using impulse and relaxation response, Ocean Eng. 28 (2001) Liu, Z., Amdahl, J. & Løset, S., Integrated numerical analysis of an iceberg collision with a foreship Struc., Marine Struc.s, 24 (2011) Hong-yu, C. & De-you, Z., A new grey neural network and its application to vibration control of offshore platform, J. of Dalian University of Technology, (2010). 50 Ya-jun, Z., De-you, Z. & Jun, M., Active control of offshore platforms using neural networks, Chuan Bo Li Xue/J. of Ship Mechanics, 7 (2003) Cui, H.Y. & Zhao, D.Y., Active control of offshore platforms using rough neural networks, Harbin Gongcheng Daxue Xuebao/J. of Harbin Eng. University, 28 (2007) Hong-yu, C., De-you, Z. & Ping, Z., Adaptive Predictive Inverse Control of Offshore Jacket Platform Based on Rough Neural Network, China Ocean Eng., 2 (2009) 185~ Kim, D.H., Kim, D. & Chang, S., Application of lattice probabilistic neural network for active response control of offshore structures, Structural Eng. and Mechanics, 31 (2009) Wei, W. & Gong-you, T., Feedback and Feedforward Optimal Control for Offshore Jacket Platforms, Oceanologia Et Limnologia Sinica, 18 (2004) Hong-nan, L. & Lin-sheng, H., Semi-Active TLCD Control of Fixed Offshore Platforms Using Artifical Neural Networks, China Ocean Eng., 17 (2003). 56 Ya-jun, Z. & De-you, Z., Neural Network-Based Active Control for Offshore Platforms, China Ocean Eng., 17 (2003) Tripathi, K.C., Rai, S. & Pandey, A.C., Southern Indian Ocean SST indices as early predictors of Indian summer monsoon, Indian J. of Marine Sci., 37 (2008) Uddin, M.A., Jameel, M., Razak, H., A. & Islam, A.B.M., Response Prediction of Offshore Floating Structure using Artificial Neural Network, Advanced Science Letters, 14 (2012) Yun, C.B. & Bahng, E.Y., Substructural identification of offshore structures using neural networks, ISOPE, GOLDEN, CO(USA), 1997, 4 (1997) Mandal, S., hegde, G. & Gupta, K.G., Evaluation of Stress Resultant of Offshore Jacket Platform using Neural Network, in Proceeding of National Seminar on Construction Management: Latest Trends and Developments Pune, India, (2004): Yasseri, S., Bahai, H., Bazargan, H. & Aminzadeh, A., Prediction of safe sea-state using finite element method and artificial neural networks, Ocean Eng., 37 (2010)