Parameters identification for ship motion model based on particle swarm optimization

Abstract

PurposeThe purpose of this paper is to identify the Nomoto ship model parameters accurately, in order to produce a very close match between the predictions based on the model and the full‐scale trials.Design/methodology/approachVarious ship maneuvering mathematical models have been used when describing the ship dynamics behavior. The Nomoto ship model is a class of simplified hydrodynamic derivative type models which are the most widely used, accepted and perhaps well developed. To determine the model parameters accurately, particle swarm optimization (PSO) is chosen as an evolution algorithm in this paper. This arithmetic can guarantee the convergence and global optimization ability, and avoid sinking into a local optimal solution.FindingsThe process of PSO for identifying the Nomoto ship model parameters is given.Research limitations/implicationsAvailability of the full‐scale trial data are the main limitations.Practical implicationsThe ship model parameters provide very useful advice in ship's autopilot process.Originality/valueThe paper presents a new parameter identification method for the second‐order Nomoto ship model based on PSO.

Combined fitness function based particle swarm optimization algorithm for system identification

“…Sum of square error (SSE) (Al-Duwaish, 2011;Alfi & Fateh, 2011;Chen et al, 2010;Yang, Maginu, & Nomura, 2009), mean square error (MSE) (Luitel & Venayagamoorthy, 2010;Nanda, Panda, & Majhi, 2010;Quaranta, Monti, & Marano, 2010;Upadhyay et al, 2014) and the root mean square error (RMSE) (Xian, Long, Li, & Wang, 2014) are usually used to determine the fitness values. They are defined as following:…”

“…The influence which acceleration coefficients c 1 ; c 2 have on the IPSO performance is tested with one benchmark IIR system 1:25z À1 À0:25z À2 1À0:3z À1 þ0:4z À2 (Luitel & Venayagamoorthy, 2010). As Fan and Zaharaa (2007) suggested to choose c 1 ; c 2 from within the interval [0.2, 2], six different acceleration coefficients are chosen as following: c 1 ¼ 2:0 and c 2 ¼ 2:0 (Xi et al, 2008), c 1 ¼ 2:0 and c 2 ¼ 1:8 (Chen et al, 2010), c 1 ¼ 1:8 and c 2 ¼ 1:8 (Wu et al, 2013), c 1 ¼ 1:49445 and c 2 ¼ 1:49445 (Tungadio et al, 2015), c 1 ¼ 0:5 þ logð2Þ and c 2 ¼ 0:5 þ logð2Þ (Yang et al, 2009), c 1 ¼ 0:2 and c 2 ¼ 0:3 (Galewski, 2014). Simulations are carried out in MATLAB, and iter max ¼ 1000, population size is 150.…”

“…It was firstly introduced by Kennedy and Eberhart , and can be easily programmed with basic mathematical and logic operation (Tungadio, Numbi, Siti, & Jimoh, 2015). In many identification applications, PSO has been successfully utilized, such as infinite impulse response (IIR) system identification problem (Upadhyay, Kar, Mandal, & Ghoshal, 2014), power system state estimation (Tungadio et al, 2015), Wiener model identification (Tang et al, 2010), and ship motion model identification (Chen, Song, & Chen, 2010).…”

Parameter Identifiability of Ship Manoeuvring Modeling Using System Identification

“…On the other hand, several novel and interesting SI methods have been proposed. Representative examples are the artificial intelligence techniques including PSO [19,20], GA [21,22], SVM [23,24], and ANN [29][30][31]. However, it should be noted that much work (including the work on classical SI or the work on new SI) concerned little about the problem of parameter identifiability especially for the effect of parameter drift or avoided dealing with the problem by selecting a simple manoeuvring model (e.g., the Nomoto model or the response model) in which the parameter drift is weak because few parameters are involved.…”

“…During last decades, the SI based method has made good progresses and many techniques have been developed in the application to ship manoeuvring modeling. Examples are least squares (LS) regression [7,8], model reference method (MRM) [9,10], extend Kalman filter (EKF) [11,12], maximum likelihood (ML) estimation [13,14], recursive prediction error (RPE) method [15,16], frequency spectrum analysis (FSA) method [17,18], particle swarm optimization (PSO) [19,20], genetic algorithm (GA) [21,22], and support vector regression (SVR) [23,24]. Generally, there have been many SI applications to the parameter identification of ship manoeuvring models.…”