An improved least squares (LS) channel estimation method based on CNN for OFDM systems

Hua Yang; Xuan Geng; Heng Xu; Yichun Shi; Hua Yang; Xuan Geng; Heng Xu; Yichun Shi

doi:10.3934/era.2023294

Electronic Research Archive

2023, Volume 31, Issue 9: 5780-5792. doi: 10.3934/era.2023294

Previous Article Next Article

Research article Special Issues

An improved least squares (LS) channel estimation method based on CNN for OFDM systems

College of Information Engineering, Shanghai Maritime University, 1550 Haigang Avenue, Pudong District, Shanghai 201306, China

Received: 31 March 2023 Revised: 07 August 2023 Accepted: 20 August 2023 Published: 28 August 2023

Least squares (LS) is a commonly used pilot-based channel estimation algorithm in orthogonal frequency division multiplexing (OFDM) systems. This algorithm is simple and easy to implement because of its low computation complexity. However, it has poor performance, especially at low signal-to-noise ratio (SNR). To solve this problem, an improved LS channel estimation method based on convolutional neural network (CNN) is proposed on the basis of analyzing the traditional LS channel estimation methods. A channel estimation compensated network is designed based on CNN, which can solve the problem of performance degradation of the mean square error (MSE) through the online and offline modules. By designing the input-output relations, training data set, and testing data set, a CNN network is iteratively trained to learn the relevant features of the channels, so that the traditional LS estimation value can be corrected to improve the accuracy. Simulation results show that the proposed method can achieve better performance in bit error rate (BER) and MSE, compared with the traditional channel estimation methods.

Keywords:

Citation: Hua Yang, Xuan Geng, Heng Xu, Yichun Shi. An improved least squares (LS) channel estimation method based on CNN for OFDM systems[J]. Electronic Research Archive, 2023, 31(9): 5780-5792. doi: 10.3934/era.2023294

Related Papers:

[1]	Zhoujin Cui . Primary resonance and feedback control of the fractional Duffing-van der Pol oscillator with quintic nonlinear-restoring force. AIMS Mathematics, 2023, 8(10): 24929-24946. doi: 10.3934/math.20231271
[2]	Zhoujin Cui, Xiaorong Zhang, Tao Lu . Resonance analysis and time-delay feedback controllability for a fractional horizontal nonlinear roller system. AIMS Mathematics, 2024, 9(9): 24832-24853. doi: 10.3934/math.20241209
[3]	P. Pirmohabbati, A. H. Refahi Sheikhani, H. Saberi Najafi, A. Abdolahzadeh Ziabari . Numerical solution of full fractional Duffing equations with Cubic-Quintic-Heptic nonlinearities. AIMS Mathematics, 2020, 5(2): 1621-1641. doi: 10.3934/math.2020110
[4]	Huitzilin Yépez-Martínez, Mir Sajjad Hashemi, Ali Saleh Alshomrani, Mustafa Inc . Analytical solutions for nonlinear systems using Nucci's reduction approach and generalized projective Riccati equations. AIMS Mathematics, 2023, 8(7): 16655-16690. doi: 10.3934/math.2023852
[5]	Ridwanulahi Iyanda Abdulganiy, Shiping Wen, Yuming Feng, Wei Zhang, Ning Tang . Adapted block hybrid method for the numerical solution of Duffing equations and related problems. AIMS Mathematics, 2021, 6(12): 14013-14034. doi: 10.3934/math.2021810
[6]	Ali Khalouta, Abdelouahab Kadem . A new computational for approximate analytical solutions of nonlinear time-fractional wave-like equations with variable coefficients. AIMS Mathematics, 2020, 5(1): 1-14. doi: 10.3934/math.2020001
[7]	Ramazan Yazgan . An analysis for a special class of solution of a Duffing system with variable delays. AIMS Mathematics, 2021, 6(10): 11187-11199. doi: 10.3934/math.2021649
[8]	Xiaoguang Yuan, Quan Jiang, Zhidong Zhou, Fengpeng Yang . The method of fundamental solutions for analytic functions in complex analysis. AIMS Mathematics, 2022, 7(4): 6820-6851. doi: 10.3934/math.2022380
[9]	Anh Tuan Nguyen, Le Dinh Long, Devendra Kumar, Van Thinh Nguyen . Regularization of a final value problem for a linear and nonlinear biharmonic equation with observed data in $L^{q}$ space. AIMS Mathematics, 2022, 7(12): 20660-20683. doi: 10.3934/math.20221133
[10]	Khalid Khan, Amir Ali, Muhammad Irfan, Zareen A. Khan . Solitary wave solutions in time-fractional Korteweg-de Vries equations with power law kernel. AIMS Mathematics, 2023, 8(1): 792-814. doi: 10.3934/math.2023039

Abstract

1. Introduction

Nonlinear oscillators have been widely used in various engineering and applied sciences, such as mathematics, physics, structural dynamics, mechanical engineering and other related fields of science ^{[1,2,3,4,5,6]}. Nonlinear differential equations (NDEs) can model many phenomena in various scientific aspects to present their effects and behaviors through mathematical principles. Perturbation methods are extremely beneficial when the nonlinear response is small ^[7,8,9,10]. In general, solving strongly nonlinear differential equations is very difficult. In ^[11], Mickens suggested an approximate expression for solving a truly nonlinear Duffing oscillator. Recently, various powerful analytical and numerical approximation techniques have been suggested for dealing with nonlinear oscillator differential equations. These include He's frequency-amplitude formulation ^[12], the harmonic balance method ^[13,14], the straightforward frequency prediction method ^[15], the modified harmonic balance method ^[16,17], the energy balance method ^[18,19,20], the homotopy perturbation method ^{[21,22,23,24,25]}, the Hamiltonian approach ^[26,27], the weighted averaging method ^[28], the global residue harmonic balance method ^[29,30,31], the max-min approach ^[32,33], Newton's harmonic balance method ^[34], the variational iteration method ^[35], the parameter-expansion method ^[36], the Lindstedt-Poincaré method ^[37,38] and the global error minimization method ^{[39,40,41,42]}.

The global error minimization method (GEMM) is one of the most frequently used techniques for dealing with the solutions of nonlinear oscillators, as it provides more accurate results valid for both weakly and strongly nonlinear oscillators than other known methods ^{[39,40,41,42]}. In the previous example, the solution up to the first approximation is calculated.

In this study, we extend and improve the global error minimization method up to a third order approximation to achieve the higher-order analytical solution of strongly nonlinear Duffing-harmonic oscillators. The present method is applied for two different problems, and the analytical results show that the modified global error minimization method (MGEMM) has better agreement with numerical solutions than the other analytical methods. Excellent agreement is observed between the approximate and exact solutions even for large amplitudes of the oscillation. Comparing the exact solutions with the approximate results has proved that the MGEMM is quite an accurate method in strongly nonlinear oscillator systems.

2. Basic idea of the modified global error minimization method (MGEMM)

To describe the proposed modified global error minimization method (MGEMM), we consider general second order nonlinear oscillator differential equations as follows:

$\ddot u + F(\dot u, \, u, \, t) = 0, \, \;\;\;u(0) = A, \, \;\;\;\dot u(0) = 0.\;\;\;\;$

(1)

We introducing $E(u)$ as a new function, defined as in ^[40,43], in the following form:

$E(u) = \int\nolimits_0^T \;{\left( {\ddot u + F(\dot u, \, u, \, t)} \right)^2}dt, \, \;\;\;\;\;T = 2\pi {\omega ^{ - 1}}.$

(2)

By assuming that $F(u)$ is an odd function, a general n-th order trial function of Eq (1) can be expressed as a sum of trigonometric functions as follows:

$u(t) = \mathop \sum \limits_{n = 0}^\infty ({a_{2n + 1)}}\cos ((2n + 1)\omega t)),$

(3)

where ${a_{(2n + 1)}}$ are unknown constant values which satisfy the relation

$A = \mathop \sum \limits_{n = 0}^\infty {a_{(2n + 1)}}.$

(4)

The following conditions were used to obtain the unknown parameters (i.e., ${a_{(2n + 1)}}$ and $\omega$ ):

$\frac{{\partial E(u)}}{{\partial \omega }} = 0, \, \;\;\;\;\;\;\;\;\frac{{\partial E(u)}}{{\partial {a_{(2n + 1)}}}} = 0, \, \;\;\;n \geqslant 1.\;\;$

(5)

By solving the n Eq (5) with the aid of Eq (4), the constants ${a_1}, \, \, {a_3}, {a_5}$ and the frequency of vibration $\omega$ are obtained.

3. Examples of two nonlinear Duffing-harmonic oscillators

In this section, two practical examples of nonlinear Duffing-harmonic oscillators are illustrated to show the effectiveness, accuracy and applicability of the proposed approach.

3.1. Example 1

In this application, we consider the following nonlinear Duffing-harmonic oscillator:

$\ddot u + {k_1}u + \frac{{{k_3}u}}{{1 + {k_2}{u^2}}} = 0, \, \;\;\;\;\;u(0) = A, \, \;\;\;\dot u(0) = 0, \,$

(6)

where dots denote differentiation with respect to $t$ . Now, we shall study some different relevant cases considering Eq (6).

3.1.1. Case 1

First, we consider ${k_1} = 1$ , ${k_2} = 1$ and ${k_3} = 1$ in Eq (6). Then, we have a nonlinear oscillator system having an irrational elastic item ^[12,16].

$\ddot u + u + \frac{u}{{1 + {u^2}}} = 0, \, \;\;\;\;\;u(0) = A, \, \;\;\;\;\;\dot u(0) = 0.$

(7)

According to the basic idea of the global error minimization method, the minimization problem of Eq (7) is

$E(u) = \int\nolimits_0^T {\left( {\ddot u + u + \frac{u}{{1 + {u^2}}}} \right)^2}dt, \, \;\;\;T = 2\pi /\omega .$

(8)

The first-order approximate solution for Eq (7) can be represented as a trial function in the form

${u_1}(t) = {a_1}\cos (\omega t).$

(9)

Substituting Eq (9) into Eq (8) and choosing ${a_1} = A,$ it follows that

$E({u_1}) = \frac{{4{A^2}\pi }}{\omega } + \frac{{3{A^4}\pi }}{\omega } + \frac{{5{A^6}\pi }}{{8\omega }} - 4{A^2}\pi \omega - \frac{9}{2}{A^4}\pi \omega - \frac{5}{4}{A^6}\pi \omega + {A^2}\pi {\omega ^3} + \frac{3}{2}{A^4}\pi {\omega ^3} + \frac{5}{8}{A^6}\pi {\omega ^3} = 0.$

(10)

Applying $\partial E({u_1})/\partial \omega = 0$ , the frequency of the nonlinear oscillator is obtained as follows:

$\omega = {\omega _1} = \sqrt {\frac{{16 + 18{A^2} + 5{A^4} + 2\sqrt {256 + 576{A^2} + 487{A^4} + 180{A^6} + 25{A^8}} }}{{3(8 + 12{A^2} + 5{A^4})}}} .$

(11)

In order to illustrate the capacity of the global error minimization method, the second-order approximation is applied to the Duffing-harmonic oscillator by using the following new trial solution:

${u_2}(t) = {a_1}\cos (\omega t) + {a_3}\cos (3\omega t), \,$

(12)

where $A = {a_1} + {a_3}.$ By substituting Eq (12) into Eq (8), we have

$\left. {\begin{array}{*{20}{l}} {E({u_2}) = \tfrac{\pi }{{8\omega }}\left( {5{{({\omega ^2} - 1)}^2}a_1^6 + 5(11{\omega ^4} - 14{\omega ^2} + 3)a_1^5{a_3} + 2a_1^3{a_3}(8(2 - 9{\omega ^2} + 5{\omega ^4})} \right.} \\ { + \left. {3(5 - 50{\omega ^2} + 109{\omega ^4})a_3^2} \right) + 3a_1^4\left( {4(2 - 3{\omega ^2} + {\omega ^4}) + (15 - 110{\omega ^2} + 159{\omega ^4})a_3^2} \right)} \\ { + a_3^2\left( {8{{(2 - 9{\omega ^2})}^2} + 12(2 - 27{\omega ^2} + 81{\omega ^4})a_3^2 + 5{{(1 - 9{\omega ^2})}^2}a_3^4} \right)} \\ { + \left. {a_1^2\left( {8{{( - 2 + {\omega ^2})}^2} + 16(6 - 45{\omega ^2} + 59{\omega ^4})a_3^2 + 3(15 - 190{\omega ^2} + 559{\omega ^4})a_3^4} \right)} \right) = 0.} \end{array}} \right)\,$

(13)

Setting $\partial E({u_2})/\partial \omega = 0$ and $\partial E({u_2})/\partial {a_3} = 0$ leads to

$\left. {\begin{array}{*{20}{l}} {\tfrac{\pi }{{8\omega }}(20\omega ( - 1 + {\omega ^2})a_1^6 + 5( - 28\omega + 44{\omega ^3})a_1^5{a_3} + 2a_1^3{a_3}(8( - 18\omega + 20{\omega ^3}) + 3( - 100\omega + 436{\omega ^3})a_3^2} \\ { + 3a_1^4(4( - 6\omega + 4{\omega ^3}) + ( - 220\omega + 636{\omega ^3})a_3^2) + a_3^2( - 288\omega (2 - 9{\omega ^2}) + 12( - 54\omega + 324{\omega ^3})a_3^2} \\ { - 180\omega (1 - 9{\omega ^2})a_3^4) + a_1^2(32\omega ( - 2 + {\omega ^2}) + 16( - 90\omega + 236{\omega ^3})a_3^2 + 3( - 380\omega + 2236{\omega ^3})a_3^4))} \\ { - \tfrac{1}{{8{\omega ^2}}}\pi (5{{( - 1 + {\omega ^2})}^2}a_1^6 + 5(3 - 14{\omega ^2} + 11{\omega ^4})a_1^5{a_3} + 2a_1^3{a_3}(8(2 - 9{\omega ^2} + 5{\omega ^4}) + 3\left( {5 - 50{\omega ^2}} \right.} \\ {\left. { + 109{\omega ^4}} \right)\, a_3^2) + 3a_1^4(4(2 - 3{\omega ^2} + {\omega ^4}) + (15 - 110{\omega ^2} + 159{\omega ^4})a_3^2) + a_3^2(8{{(2 - 9{\omega ^2})}^2}} \\ { + 12(2 - 27{\omega ^2} + 81{\omega ^4})a_3^2 + 5{{(1 - 9{\omega ^2})}^2}a_3^4) + a_1^2(8{{( - 2 + {\omega ^2})}^2} + 16(6 - 45{\omega ^2} + 59{\omega ^4})a_3^2} \\ { + 3(15 - 190{\omega ^2} + 559{\omega ^4})a_3^4)) = 0, } \end{array}} \right)\,$

(14)

$\left. {\begin{array}{*{20}{l}} {\tfrac{\pi }{{8\omega }}(5(3 - 14{\omega ^2} + 11{\omega ^4})a_1^5 + 6(15 - 110{\omega ^2} + 159{\omega ^4})a_1^4{a_3} + 12(5 - 50{\omega ^2} + 109{\omega ^4})a_1^3a_3^2} \\ { + 2a_1^3\left( {8(2 - 9{\omega ^2} + 5{\omega ^4}) + 3(5 - 50{\omega ^2} + 109{\omega ^4})a_3^2} \right) + a_3^2\left( {24(2 - 27{\omega ^2}81{\omega ^4}){a_3} \\+ 20{{(1 - 9{\omega ^2})}^2}a_3^3} \right)} \\ { + a_1^2\left( {32(6 - 45{\omega ^2} + 59{\omega ^4}){a_3} + 12(15 - 190{\omega ^2} + 559{\omega ^4})a_3^3} \right) + 2{a_3}\left( {8{{(2 - 9{\omega ^2})}^2}} \right.} \\ { + \left. {12(2 - 27{\omega ^2} + 81{\omega ^4})a_3^2 + 5{{(1 - 9{\omega ^2})}^2}a_3^2)} \right) = 0.} \end{array}} \right)\,$

(15)

For a known amplitude, the parameters of ${a_1}$ , ${a_3}$ and angular frequency $\omega$ can be obtained by using the condition $A = {a_1} + {a_3}$ and solving Eqs (14) and (15). The computations were performed using the Mathematica software program, version 9.

To illustrate the capacity of this method, the third order approximation is applied by using the following trial function:

${u_3}(t) = {a_1}\cos (\omega t) + {a_3}\cos (3\omega t) + {a_5}\cos (5\omega t), \,$

(16)

where $A = {a_1} + {a_3} + {a_5}$ . Bringing Eq (16) into Eq (8) results in

$\left. \begin{gathered} \begin{array}{*{20}{l}} {E({u_3}) = \, \tfrac{\pi }{{8\omega }}\left( {5a_1^6} \right.{{\left( {{\omega ^2} - 1} \right)}^2} + a_1^5\left( {{\omega ^2} - 1} \right)\, \left( {5{a_3}\left( {11{\omega ^2} - 3} \right) + 3{a_5}\left( {9{\omega ^2} - 1} \right)} \right) + 5a_3^6{{\left( {1 - 9{\omega ^2}} \right)}^2}} \\ { + 2a_1^3{a_3}\left( {3a_3^2\left( {109{\omega ^4} - 50{\omega ^2} + 5} \right) + 3{a_5}{a_3}\left( {663{\omega ^4} - 230{\omega ^2} + 15} \right) + 8\left( {5{\omega ^4} - 9{\omega ^2} + 2} \right)} \right.} \\ {\left. { + 10a_5^2\left( {251{\omega ^4} - 62{\omega ^2} + 3} \right)} \right) + 3a_3^4\left( {a_5^2\left( {2911{\omega ^4} - 430{\omega ^2} + 15} \right) + 4\left( {81{\omega ^4} - 27{\omega ^2} + 2} \right)} \right)} \\ { + 2{a_1}a_3^2{a_5}\left( {a_3^2\left( {2894{\omega ^4} - 620{\omega ^2} + 30} \right) + 15{a_5}{a_3}\left( {153{\omega ^4} - 26{\omega ^2} + 1} \right) + 8\left( {287{\omega ^4} - 99{\omega ^2} + 6} \right)} \right.} \\ {\left. { + 3a_5^2\left( {3399{\omega ^4} - 470{\omega ^2} + 15} \right)} \right) + a_5^2\left( {5a_5^4{{\left( {1 - 25{\omega ^2}} \right)}^2} + 12a_5^2\left( {625{\omega ^4} - 75{\omega ^2} + 2} \right) \\+ 8{{\left( {2 - 25{\omega ^2}} \right)}^2}} \right)} \\ { + a_3^2\left( {3a_5^4\left( {5631{\omega ^4} - 590{\omega ^2} + 15} \right) + 16a_5^2\left( {803{\omega ^4} - 153{\omega ^2} + 6} \right) + 8{{\left( {2 - 9{\omega ^2}} \right)}^2}} \right)} \end{array} \hfill \\ \begin{array}{*{20}{l}} { + a_3^2{{\left( {3a_5^4\left( {5631{\omega ^4} - 590{\omega ^2} + 15} \right) + 16a_5^2\left( {803{\omega ^4} - 153{\omega ^2} + 6} \right) + 8\left( {2 - 9{\omega ^2}} \right){\omega ^2}} \right)}^2}} \\ {\left. {\left. { + 3a_5^2\left( {277{\omega ^4} - 90{\omega ^2} + 5} \right)} \right)} \right) + a_1^2\left( {3a_3^4\left( {559{\omega ^4} - 190{\omega ^2} + 15} \right) + 18{a_5}a_3^3\left( {341{\omega ^4} - 90{\omega ^2} + 5} \right)} \right.} \\ {\left. { + 8{{\left( {{\omega ^2} - 2} \right)}^2}} \right) + 9a_5^4\left( {1317{\omega ^4} - 170{\omega ^2} + 5} \right) + 48a_5^2\left( {121{\omega ^4} - 39{\omega ^2} + 2} \right)} \\ {4a_3^2\left( {3a_5^2\left( {1743{\omega ^4} - 350{\omega ^2} + 15} \right) + 4\left( {59{\omega ^4} - 45{\omega ^2} + 6} \right)} \right)} \\ {\left. {\left. { + 6{a_3}{a_5}\left( {a_5^2\left( {2719{\omega ^4} - 430{\omega ^2} + 15} \right) + 8\left( {49{\omega ^4} - 27{\omega ^2} + 2} \right)} \right)} \right)} \right) = 0.} \end{array} \hfill \\ \end{gathered} \right)\,$

(17)

Applying $\partial E({u_3})/\partial \omega = 0$ , $\partial E({u_3})/\partial {a_3} = 0$ and $\partial E({u_3})/\partial {a_5} = 0$ yields

$\left. {\begin{array}{*{20}{l}} {\tfrac{\pi }{{8\omega }}\left( {20a_1^6\omega \left( {{\omega ^2} - 1} \right) + a_1^5\left( {{\omega ^2} - 1} \right)\, \left( {110{a_3}\omega + 54{a_5}\omega } \right) + 2a_1^5\omega \left( {5{a_3}\left( {11{\omega ^2} - 3} \right) + 3{a_5}\left( {9{\omega ^2} - 1} \right)} \right)} \right.} \\ {\left. { + 3{a_5}\left( {9{\omega ^2} - 1} \right)} \right) + 2a_1^3{a_3}\left( {3a_3^2\left( {436{\omega ^3} - 100\omega } \right) + 3{a_5}{a_3}\left( {2652{\omega ^3} - 460\omega } \right) + 8\left( {20{\omega ^3} - 18\omega } \right)} \right.} \\ {\left. { + 10a_5^2\left( {1004{\omega ^3} - 124\omega } \right)} \right) + 3a_3^4\left( {a_5^2\left( {11644{\omega ^3} - 860\omega } \right) + 4\left( {324{\omega ^3} - 54\omega } \right)} \right)} \\ { + 2{a_1}a_3^2{a_5}\left( {a_3^2\left( {11576{\omega ^3} - 1240\omega } \right) + 15{a_5}{a_3}\left( {612{\omega ^3} - 52\omega } \right) + 8\left( {1148{\omega ^3} - 198\omega } \right)} \right.} \\ {\left. { + 3a_5^2\left( {13596{\omega ^3} - 940\omega } \right)} \right) + a_5^2\left( {12a_5^2\left( {2500{\omega ^3} - 150\omega } \right) - 800\omega \left( {2 - 25{\omega ^2}} \right)} \right.} \\ {\left. { - 500a_5^4\omega \left( {1 - 25{\omega ^3}} \right)} \right) + a_3^2\left( {3a_5^4\left( {22524{\omega ^3} - 1180\omega } \right) + 16a_5^2\left( {3212{\omega ^3} - 306\omega } \right) \\- 288\omega \left( {2 - 9{\omega ^2}} \right)} \right)} \\ \begin{gathered} + a_1^4\left( {a_3^2\left( {1908{\omega ^3} - 660\omega } \right) + 4{a_5}{a_3}\left( {1468{\omega ^3} - 380\omega } \right) + 3\left( {3a_5^2\left( {1108{\omega ^3} - 180\omega } \right) \\+ 4\left( {4{\omega ^3} - 6\omega } \right)} \right)} \right) \hfill \\ \begin{array}{*{20}{l}} {a_1^2\left( {3a_3^4\left( {2236{\omega ^3} - 380\omega } \right) + 18{a_5}a_3^3\left( {1364{\omega ^3} - 180\omega } \right) + 48a_5^2\left( {484{\omega ^3} - 78\omega } \right) \\+ 32\omega \left( {{\omega ^2} - 2} \right)} \right.} \\ { + 9a_5^4\left( {5268{\omega ^3} - 340\omega } \right) + 4a_3^2\left( {3a_5^2\left( {6972{\omega ^3} - 700\omega } \right) + 4\left( {236{\omega ^3} - 90\omega } \right)} \right)} \\ {\left. {\left. { + 6{a_3}{a_5}\left( {a_5^2\left( {10876{\omega ^3} - 860\omega } \right) + 8\left( {196{\omega ^3} - 54\omega } \right)} \right)} \right)} \right)} \\ { - \tfrac{\pi }{{8{\omega ^2}}}\left( {5a_1^6{{\left( {{\omega ^2} - 1} \right)}^2} + a_1^5\left( {{\omega ^2} - 1} \right)\, \left( {5{a_3}\left( {11{\omega ^2} - 3} \right) + 3{a_5}\left( {9{\omega ^2} - 1} \right)} \right) + 5a_3^6{{\left( {1 - 9{\omega ^2}} \right)}^2}} \right.} \\ { + 2a_1^3{a_3}\left( {3a_3^2\left( {109{\omega ^4} - 50{\omega ^2} + 5} \right) + 3{a_5}{a_3}\left( {663{\omega ^4} - 230{\omega ^2} + 15} \right) + 8\left( {5{\omega ^4} - 9{\omega ^2} + 2} \right)} \right.} \\ {\left. { + 10a_5^2\left( {251{\omega ^4} - 62{\omega ^2} + 3} \right)} \right) + 3a_3^4\left( {a_5^2\left( {2911{\omega ^4} - 430{\omega ^2} + 15} \right) + 4\left( {81{\omega ^4} - 27{\omega ^2} + 2} \right)} \right)} \\ { + 2{a_1}a_3^2{a_5}\left( {a_3^2\left( {2894{\omega ^4} - 620{\omega ^2} + 30} \right) + 15{a_5}{a_3}\left( {153{\omega ^4} - 26{\omega ^2} + 1} \right) \\+ 8\left( {287{\omega ^4} - 99{\omega ^2} + 6} \right)} \right.} \end{array} \hfill \\ \begin{array}{*{20}{l}} {\left. { + 3a_5^2\left( {3399{\omega ^4} - 470{\omega ^2} + 15} \right)} \right) + a_5^2\left( {5a_3^4{{\left( {1 - 25{\omega ^2}} \right)}^2} + 12a_5^2\left( {625{\omega ^4} - 75{\omega ^2} + 2} \right) \\+ 8{{\left( {2 - 25{\omega ^2}} \right)}^2}} \right)} \\ { + a_3^2\left( {3a_5^4\left( {5631{\omega ^4} - 590{\omega ^2} + 15} \right) + 16a_5^2\left( {803{\omega ^4} - 153{\omega ^2} + 6} \right) + 8{{\left( {2 - 9{\omega ^2}} \right)}^2}} \right)} \\ { + a_3^2\left( {3a_5^4\left( {5631{\omega ^4} - 590{\omega ^2} + 15} \right) + 16a_5^2\left( {803{\omega ^4} - 153{\omega ^2} + 6} \right) + 8{{\left( {2 - 9{\omega ^2}} \right)}^2}} \right)} \\ {\left. {\left. { + 3a_5^2\left( {277{\omega ^4} - 90{\omega ^2} + 5} \right)} \right)} \right) + a_1^2\left( {3a_3^4\left( {559{\omega ^4} - 190{\omega ^2} + 15} \right) + 8{{\left( {{\omega ^2} - 2} \right)}^2}} \right.} \\ { + 9a_5^4\left( {1317{\omega ^4} - 170{\omega ^2} + 5} \right) + 48a_5^2\left( {121{\omega ^4} - 39{\omega ^2} + 2} \right) + 18a_3^3{a_5}\left( {341{\omega ^4} - 90{\omega ^2} + 5} \right)} \\ { + 4a_3^2\left( {3a_5^2\left( {1743{\omega ^4} - 350{\omega ^2} + 15} \right) + 4\left( {59{\omega ^4} - 45{\omega ^2} + 6} \right)} \right)} \\ {\left. {\left. { + 6{a_3}{a_5}\left( {a_5^2\left( {2719{\omega ^4} - 430{\omega ^2} + 15} \right) + 8\left( {49{\omega ^4} - 27{\omega ^2} + 2} \right)} \right)} \right)} \right) = 0, } \end{array} \hfill \\ \end{gathered} \end{array}} \right)\,$

(18)

$\left. \begin{gathered} \begin{array}{*{20}{l}} {\, \tfrac{\pi }{{8\omega }}\left( {5a_1^5} \right.\left( {{\omega ^2} - 1} \right)\, \left( {11{\omega ^2} - 3} \right) + 30a_3^5{{\left( {1 - 9{\omega ^2}} \right)}^2} + 2{a_1}a_3^2{a_5}\left( {2{a_3}\left( {2894{\omega ^4} - 620{\omega ^2} + 30} \right)} \right.} \\ {\left. { + 15{a_5}\left( {153{\omega ^4} - 26{\omega ^2} + 1} \right)} \right) + a_1^4\left( {2{a_3}\left( {477{\omega ^4} - 330{\omega ^2} + 45} \right) + 4{a_5}\left( {367{\omega ^4} - 190{\omega ^2} + 15} \right)} \right)} \\ { + 2a_1^3{a_3}\left( {6{a_3}\left( {109{\omega ^4} - 50{\omega ^2} + 5} \right) + 3{a_5}\left( {663{\omega ^4} - 230{\omega ^2} + 15} \right)} \right) + 2a_1^3\left( {8\left( {5{\omega ^4} - 9{\omega ^2} + 2} \right)} \right.} \\ {\left. { + 3a_3^2\left( {109{\omega ^4} - 50{\omega ^2} + 5} \right) + 3{a_5}{a_3}\left( {663{\omega ^4} - 230{\omega ^2} + 15} \right) + 10a_5^2\left( {251{\omega ^4} - 62{\omega ^2} + 3} \right)} \right)} \\ { + 12a_3^3\left( {a_5^2\left( {2911{\omega ^4} - 430{\omega ^2} + 15} \right) + 4\left( {81{\omega ^4} - 27{\omega ^2} + 2} \right)} \right) + \\ 4{a_1}{a_3}{a_5}\left( {8\left( {287{\omega ^4} - 99{\omega ^2} + 6} \right)} \right.} \end{array} \hfill \\ \begin{array}{*{20}{l}} {\left. { + a_3^2\left( {2894{\omega ^4} - 620{\omega ^2} + 30} \right) + 15{a_5}{a_3}\left( {153{\omega ^4} - 26{\omega ^2} + 1} \right) + 3a_5^2\left( {3399{\omega ^4} - 470{\omega ^2} + 15} \right)} \right)} \\ { + 2{a_3}\left( {3a_5^4\left( {5631{\omega ^4} - 590{\omega ^2} + 15} \right) + 16a_5^2\left( {803{\omega ^4} - 153{\omega ^2} + 6} \right) + 8{{\left( {2 - 9{\omega ^2}} \right)}^2}} \right)} \\ { + a_1^2\left( {12a_3^3\left( {559{\omega ^4} - 190{\omega ^2} + 15} \right) + 54a_3^2{a_5}\left( {341{\omega ^4} - 90{\omega ^2} + 5} \right) + } \right.\, 8{a_3}\left( {4\left( {59{\omega ^4} - 45{\omega ^2} + 6} \right)} \right.} \\ {\left. {\left. {\left. { + 3a_5^2\left( {1743{\omega ^4} - 350{\omega ^2} + 15} \right)} \right) + 6{a_5}\left( {a_5^2\left( {2719{\omega ^4} - 430{\omega ^2} + 15} \right) + \\ 8\left( {49{\omega ^4} - 27{\omega ^2} + 2} \right)} \right)} \right)} \right) = 0, } \end{array} \hfill \\ \end{gathered} \right)\,$

(19)

$\begin{gathered} \left. \begin{gathered} \begin{array}{*{20}{l}} {\, \tfrac{\pi }{{8\omega }}\left( {3a_1^5} \right.\left( {{\omega ^2} - 1} \right)\, \left( {9{\omega ^2} - 1} \right) + 6a_3^4{a_5}\left( {2911{\omega ^4} - 430{\omega ^2} + 15} \right) + 2a_1^3{a_3}\left( {3{a_3}\left( {663{\omega ^4} - 230{\omega ^2} + 15} \right)} \right.} \\ {\left. { + 20{a_5}\left( {251{\omega ^4} - 62{\omega ^2} + 3} \right)} \right) + a_1^4\left( {4{a_3}\left( {367{\omega ^4} - 190{\omega ^2} + 15} \right) + 18{a_5}\left( {277{\omega ^4} - 90{\omega ^2} + 5} \right)} \right)} \\ { + 2{a_1}a_3^2{a_5}\left( {15{a_3}\left( {153{\omega ^4} - 26{\omega ^2} + 1} \right) + 6{a_5}\left( {3399{\omega ^4} - 470{\omega ^2} + 15} \right)} \right) + \\ 2{a_1}a_3^2\left( {8\left( {287{\omega ^4} - 99{\omega ^2} + 6} \right)} \right.} \\ { + \left. {a_3^2\left( {2894{\omega ^4} - 620{\omega ^2} + 30} \right) + 15{a_5}{a_3}\left( {153{\omega ^4} - 26{\omega ^2} + 1} \right) + 3a_5^2\left( {3399{\omega ^4} - 470{\omega ^2} + 15} \right)} \right)} \\ { + a_5^2\left( {20a_5^3{{\left( {1 - 25{\omega ^2}} \right)}^2} + 24{a_5}\left( {625{\omega ^4} - 75{\omega ^2} + 2} \right)} \right) + a_3^2\left( {32{a_5}\left( {803{\omega ^4} - 153{\omega ^2} + 6} \right)} \right.} \end{array} \hfill \\ \begin{array}{*{20}{l}} {\left. { + 12a_5^3\left( {5631{\omega ^4} - 590{\omega ^2} + 15} \right)} \right) + 2{a_5}\left( {12a_5^2\left( {625{\omega ^4} - 75{\omega ^2} + 2} \right) + 8{{\left( {2 - 25{\omega ^2}} \right)}^2}} \right.} \\ {\left. { + 5a_5^4{{\left( {1 - 25{\omega ^2}} \right)}^2}} \right) + a_1^2\left( {18a_3^3\left( {341{\omega ^4} - 90{\omega ^2} + 5} \right) + 96{a_5}\left( {121{\omega ^4} - 39{\omega ^2} + 2} \right)} \right.} \\ { + 36a_5^3\left( {1317{\omega ^4} - 170{\omega ^2} + 5} \right) + 12{a_3}a_5^2\left( {2719{\omega ^4} - 430{\omega ^2} + 15} \right) + 24a_3^2{a_5} \\\left( {1743{\omega ^4} - 350{\omega ^2} + 15} \right)} \\ {\left. {\left. { + 6{a_3}\left( {a_5^2\left( {2719{\omega ^4} - 430{\omega ^2} + 15} \right) + 8\left( {49{\omega ^4} - 27{\omega ^2} + 2} \right)} \right)} \right)} \right) = 0.} \end{array} \hfill \\ \end{gathered} \right) \hfill \\ \, \hfill \\ \end{gathered}$

(20)

Now, by solving Eqs (18)–(20) and applying the condition $A = {a_1} + {a_3} + {a_5}$ , the parameters ${a_1}, \, {a_3}, \, {a_5}$ and the angular frequency $\omega$ can be obtained for the known amplitude $A,$ using the Mathematica software program, version 9. To examine the accuracy of the MGEMM solutions, the obtained results are compared with the frequency-amplitude formulation (FAF) ^[12], the energy balance method (EBM) ^[19], the modified harmonic balance method (MHBM) ^[16] and the exact solutions, as presented in Table 1 and Figure 1. We conclude that the third order approximation provides an excellent accuracy with respect to the exact numerical solutions.

Table 1. Comparison of the approximate analytical frequencies with the exact solutions.

$A$	${\omega _{FAF}}$	${\omega _{EBM}}$	${\omega _{MHBM}}$	${\omega _{3rd\, GEMM}}$	${\omega _{exact}}$
	^[12]	^[19]	^[16]	present	Exact
0.01	1.41419	1.41419	1.41419	1.41419	1.41419
0.1	1.41158	1.41158	1.41158	1.41158	1.41158
0.2	1.40388	1.40389	1.40390	1.40390	1.40390
0.4	1.37581	1.37595	1.37616	1.37616	1.37616
0.6	1.33694	1.33743	1.33827	1.33827	1.33827
0.8	1.29448	1.29550	1.29744	1.29743	1.29743
1	1.25375	1.25514	1.25845	1.25840	1.25842
5	1.02500	1.02588	1.03148	1.02945	1.03139
10	1.00656	1.00681	1.00895	1.00790	1.00893
100	1.00007	1.0000	1.00010	1.00010	1.00010
1000	1.00000	1.00000	1.00000	1.00000	1.00000

| Show Table

DownLoad: CSV

Figure 1. Comparison of the approximate solution (red line) with the numerical solution (blue line).

DownLoad: Full-Size Img PowerPoint

3.1.2. Case 2

Now, if we put ${k_1} = 0$ , ${k_2} = 1$ and ${k_3} = 1$ in Eq (6), we obtain the following nonlinear oscillator, in which the restoring force has a rational expression ^[21,26].

$\ddot u + \frac{u}{{1 + {u^2}}} = 0, \, \;\;\;\;\;u(0) = A, \, \;\;\;\dot u(0) = 0.$

(21)

Using the previously mentioned procedure, the solution up to a third-order approximation is calculated. Depending on the analytical approximation, first, second or third, the approximate solution is assumed in the forms of (9), (12) and (16), respectively.

Finally, as in Case 1, the third order approximate solutions are compared with the homotopy perturbation method (HPM) ^[21], the Hamiltonian approach (HA) ^[26] and the exact solutions, as displayed in Table 2.

Table 2. Comparison of the approximate analytical frequencies with the exact solutions.

$A$	${\omega _{HPM}}$	${\omega _{HA}}$	${\omega _{3rd\, GEMM}}$	${\omega _{exact}}$
	^[21]	^[26]	present	Exact
0.01	0.999963	0.9999625	0.999963	0.99999
0.1	0.996271	0.99627403	0.996273	0.9991208
1	0.755929	0.765366864	0.761539	0.76157808
10	0.114708	0.13420106	0.11948	0.123322
100	0.0115462	0.01407125	0.0120357	0.0125265

| Show Table

DownLoad: CSV

3.2. Example 2

In the second application, we will consider the following nonlinear Duffing-harmonic oscillator ^[18]:

$\ddot u + {k_1}u + \frac{{{k_3}{u^3}}}{{1 + {k_2}{u^2}}} = 0, \, \;\;\;\;\;u(0) = A, \, \;\;\;\;\;\dot u(0) = 0, \,$

(22)

where dots denote differentiation with respect to $t$ . Now, we consider some following cases to compare the present solutions with published solutions using different approximate analytical methods.

3.2.1. Case 1

First, we consider ${k_1} = 1,$ ${k_2} = 1$ and ${k_3} = 1$ in Eq (22). Then, we have the following nonlinear Duffing-harmonic equation ^[18,28]:

$\ddot u + u + \frac{{{u^3}}}{{1 + {u^2}}} = 0, \, \;\;\;\;\;u(0) = A, \, \, \, \dot u(0) = 0.$

(23)

The minimization problem is

$E(u) = \int\nolimits_0^T {\left( {\ddot u + u + \frac{{{u^3}}}{{1 + {u^2}}}} \right)^2}dt, \, \;\;T = \frac{{2\pi }}{\omega }.$

(24)

For the first-order approximation, assume that the trial function is given by

${u_1}(t) = {a_1}\cos (\omega t), \,$

(25)

where $A = {a_1}$ . By inserting Eq (25) into Eq (24), we obtain

$E({u_1}) = \frac{{{A^2}\pi }}{\omega } + \frac{{3{A^4}\pi }}{\omega } + \frac{{5{A^6}\pi }}{{2\omega }} - 2{A^2}\pi \omega - \frac{9}{2}{A^4}\pi \omega - \frac{5}{2}{A^6}\pi \omega + {A^2}\pi {\omega ^3} + \frac{3}{2}{A^4}\pi {\omega ^3} + \frac{5}{8}{A^6}\pi {\omega ^3} = 0.$

(26)

The frequency can be found through the condition $\partial E({u_1})/\partial \omega = 0$ , as follows:

$\omega = {\omega _1} = \sqrt {\frac{{8 + 18{A^2} + 10{A^4} + 2\sqrt {64 + 288{A^2} + 487{A^4} + 360{A^6} + 100{A^8}} }}{{3(8 + 12{A^2} + 5{A^4})}}} .$

(27)

To improve the analytical approximation, we add additional terms to the trial function:

${u_2}(t) = {a_1}\cos (\omega t) + {a_3}\cos (3\omega t).$

(28)

The constraint of this minimization is $A = {a_1} + {a_3}$ . Substituting the above new trial function into Eq (24), we obtain

$\left. {\begin{array}{*{20}{l}} {E({u_2}) = \, \tfrac{\pi }{{8\omega }}\left( {5a_1^6} \right.{{\left( {{\omega ^2} - 2} \right)}^2} + 5a_1^5{a_3}\left( {11{\omega ^4} - 28{\omega ^2} + 12} \right) + 3a_1^4\left( {4\left( {{\omega ^4} - 3{\omega ^2} + 2} \right)} \right.} \\ {\left. { + a_3^2\left( {159{\omega ^4} - 220{\omega ^2} + 60} \right)} \right) + 2a_1^3{a_3}\left( {a_3^2\left( {327{\omega ^4} - 300{\omega ^2} + 60} \right) + 8\left( {5{\omega ^4} - 9{\omega ^2} + 2} \right)} \right)} \\ { + a_3^2\left( {5a_3^4{{\left( {2 - 9{\omega ^2}} \right)}^2} + 12a_3^2\left( {81{\omega ^4} - 27{\omega ^2} + 2} \right) + 8{{\left( {1 - 9{\omega ^2}} \right)}^2}} \right)} \\ {\left. { + a_1^2\left( {3a_3^4\left( {559{\omega ^4} - 380{\omega ^2} + 60} \right) + 16a_3^2\left( {59{\omega ^4} - 45{\omega ^2} + 6} \right) + 8{{\left( {{\omega ^2} - 1} \right)}^2}} \right)} \right) = 0.} \end{array}} \right)\,$

(29)

By using $\partial E({u_2})/\partial \omega = 0$ and $\partial E({u_2})/\partial {a_3} = 0$ , it follows that

$\left. {\begin{array}{*{20}{l}} {\begin{array}{*{20}{l}} {\tfrac{\pi }{{8\omega }}\left( {5{a_3}a_1^5} \right.\, \left( {44{\omega ^3} - 56\omega } \right) + 3a_1^4\left( {a_3^2\left( {636{\omega ^3} - 440\omega } \right) + 4\left( {4{\omega ^3} - 6\omega } \right)} \right) + 20a_1^6\omega \left( {{\omega ^2} - 2} \right)} \end{array}} \\ { + 2a_1^3{a_3}\left( {a_3^2\left( {1308{\omega ^3} - 600\omega } \right) + 8\left( {20{\omega ^3} - 18\omega } \right)} \right) + a_3^2\left( {12a_3^2\left( {324{\omega ^3} - 54\omega } \right) - 288\omega \left( {1 - 9{\omega ^3}} \right)} \right.} \\ {\left. {\left. { - 180a_3^4\omega \left( {2 - 9{\omega ^3}} \right)} \right) + a_1^2\left( {3a_3^4\left( {2236{\omega ^3} - 760\omega } \right) + 16a_3^2\left( {236{\omega ^3} - 90\omega } \right) + 32\omega \left( {{\omega ^2} - 1} \right)} \right)} \right)} \\ { - \tfrac{\pi }{{8{\omega ^2}}}\left( {5a_1^6{{\left( {{\omega ^2} - 2} \right)}^2} + 5{a_3}a_1^5\left( {11{\omega ^4} - 28{\omega ^2} + 12} \right) + 3a_1^4\left( {a_3^2\left( {159{\omega ^4} - 220{\omega ^2} + 60} \right) + 4\left( {{\omega ^4} - 3{\omega ^2} + 2} \right)} \right)} \right.} \\ { + 2a_1^3{a_3}\left( {a_3^2\left( {327{\omega ^4} - 300{\omega ^2} + 60} \right) + 8\left( {5{\omega ^4} - 9{\omega ^2} + 2} \right)} \right) + a_3^2\left( {12a_3^2\left( {81{\omega ^4} - 27{\omega ^2} + 2} \right) + 8{{\left( {1 - 9{\omega ^2}} \right)}^2}} \right.} \\ {\left. {\left. { + 5a_3^4{{\left( {2 - 9{\omega ^2}} \right)}^2}} \right) + a_1^2\left( {3a_3^4\left( {559{\omega ^4} - 380{\omega ^2} + 60} \right) + 16a_3^2\left( {59{\omega ^4} - 45{\omega ^2} + 6} \right) + 8{{\left( {{\omega ^2} - 1} \right)}^2}} \right)} \right) = 0, } \end{array}} \right)\,$

(30)

$\left. {\begin{array}{*{20}{l}} {\tfrac{\pi }{{8\omega }}\left( {5a_1^5\left( {11{\omega ^4} - 28{\omega ^2} + 12} \right) + 6{a_3}a_1^4\left( {159{\omega ^4} - 220{\omega ^2} + 60} \right) + 4a_3^2a_1^3\left( {327{\omega ^4} - 300{\omega ^2} + 60} \right)} \right.} \\ { + 2a_1^3\left( {a_3^2\left( {327{\omega ^4} - 300{\omega ^2} + 60} \right) + 8\left( {5{\omega ^4} - 9{\omega ^2} + 2} \right)} \right) + a_3^2\left( {24{a_3}\left( {81{\omega ^4} - 27{\omega ^2} + 2} \right)} \right.} \\ {\left. { + 20a_3^3{{\left( {2 - 9{\omega ^2}} \right)}^2}} \right) + a_1^2\left( {12a_3^3\left( {559{\omega ^4} - 380{\omega ^2} + 60} \right) + 32{a_3}\left( {59{\omega ^4} - 45{\omega ^2} + 6} \right)} \right)} \\ {\left. { + 2{a_3}\left( {5a_3^4{{\left( {2 - 9{\omega ^2}} \right)}^2} + 12a_3^2\left( {81{\omega ^4} - 27{\omega ^2} + 2} \right) + 8{{\left( {1 - 9{\omega ^2}} \right)}^2}} \right)} \right) = 0.} \end{array}} \right)\,$

(31)

The minimization problem's conditions can be easily achieved by replacing ${a_1} = A - {a_3}$ , and the parameters ${a_1}$ , ${a_3}$ and angular frequency $\omega$ can be obtained for a known amplitude $A$ .

To show the accuracy of the MGEM method in higher order approximations, we apply the third order approximation and consider the following trial function:

${u_3}(t) = {a_1}\cos (\omega t) + {a_3}\cos (3\omega t) + {a_5}\cos (5\omega t).$

(32)

Using Eq (32) as the trial function in Eq (24), where $A = {a_1} + {a_3} + {a_5}$ , leads to

$\left. {\begin{array}{*{20}{l}} {E({u_3}) = \, \tfrac{{\pi \omega }}{8}\left( {5a_1^6} \right.\left( {{\omega ^2} - 4} \right) + a_1^5\left( {5{a_3}\left( {11{\omega ^2} - 28} \right) + 3{a_5}\left( {9{\omega ^2} - 20} \right)} \right) + 45a_3^6\left( {9{\omega ^2} - 4} \right)} \\ { + 2a_1^3{a_3}\left( {a_3^2\left( {327{\omega ^2} - 300} \right) + 10a_5^2\left( {251{\omega ^2} - 124} \right) + 3{a_3}{a_5}\left( {663{\omega ^2} - 460} \right) + 40{\omega ^2}} \right)} \\ { + a_1^4\left( {a_3^2\left( {477{\omega ^2} - 660} \right) + 9a_5^2\left( {277{\omega ^2} - 180} \right) + 4{a_3}{a_5}\left( {367{\omega ^2} - 380} \right) + 12{\omega ^2}} \right)} \\ { + 3a_3^4\left( {a_5^2\left( {2911{\omega ^2} - 860} \right) + 324{\omega ^2}} \right) + 2{a_1}a_3^2{a_5}\left( {2a_3^2\left( {1447{\omega ^2} - 620} \right) + 2296{\omega ^2}} \right.} \\ {\left. { + 3a_5^2\left( {3399{\omega ^2} - 940} \right) + 15{a_3}{a_5}\left( {153{\omega ^2} - 52} \right)} \right) + 125a_5^2\left( {60a_5^2{\omega ^2} + 40{\omega ^2}} \right.} \\ {\left. { + a_5^4\left( {25{\omega ^2} - 4} \right)} \right) + a_1^2\left( {3a_3^4\left( {559{\omega ^2} - 380} \right) + 944a_3^2{\omega ^2} + 8{\omega ^2}} \right.} \\ { + 6{a_3}{a_5}\left( {3a_3^2\left( {341{\omega ^2} - 180} \right) + 392{\omega ^2}} \right) + 12a_5^2\left( {7a_3^2\left( {249{\omega ^2} - 100} \right) + 484{\omega ^2}} \right)} \\ {\left. { + 9a_5^4\left( {1317{\omega ^2} - 340} \right) + 6{a_3}a_5^3\left( {2719{\omega ^2} - 860} \right)} \right) + a_3^2\left( {12848a_5^2{\omega ^2} + 648{\omega ^2}} \right.} \\ {\left. {\left. { + 3a_5^4\left( {5631{\omega ^2} - 1180} \right)} \right)} \right) = 0.} \end{array}} \right)\,$

(33)

By setting $\partial E({u_3})/\partial \omega = 0$ , $\partial E({u_3})/\partial {a_3} = 0$ and $\partial E({u_3})/\partial {a_5} = 0$ , we obtain

$\left. \begin{gathered} \begin{array}{*{20}{l}} {\tfrac{1}{{8\omega }}\pi \left( {20a_1^6\omega } \right.\left( {{\omega ^2} - 2} \right) - 180a_3^6\omega \left( {2 - 9{\omega ^2}} \right) + a_1^5\left( {{\omega ^2} - 2} \right)\, \left( {110{a_3}\omega + 54{a_5}\omega } \right) + 2{\omega _3}a_1^5\left( {5{a_3}\left( {11{\omega ^2} - 6} \right)} \right.} \\ {\left. { + 3{a_5}\left( {9{\omega ^2} - 2} \right)} \right) + 3a_3^4\left( {a_5^2\left( {11644{\omega ^3} - 1720\omega } \right) + 4\left( {324{\omega ^3} - 54\omega } \right)} \right) + 2{a_1}a_3^2{a_5}\left( {8\left( {1148{\omega ^3} - 198\omega } \right)} \right.} \\ {\left. { + 2a_3^2\left( {5788{\omega ^3} - 1240\omega } \right) + 15{a_5}{a_3}\left( {612{\omega ^3} - 104\omega } \right) + 3a_5^2\left( {13596{\omega ^3} - 1880\omega } \right)} \right) + a_5^2\left( { - 800\omega \left( {1 - 25{\omega ^2}} \right)} \right.} \\ {\left. { + 12a_5^2\left( {2500{\omega ^3} - 150\omega } \right) - 500a_5^4\omega \left( {2 - 25{\omega ^2}} \right)} \right) + a_3^2\left( {16a_5^2\left( {3212{\omega ^3} - 306\omega } \right) - 288\omega \left( {1 - 9{\omega ^2}} \right)} \right.} \\ {\left. { + 3a_5^4\left( {22524{\omega ^3} - 2360\omega } \right)} \right) + 2a_1^3{a_3}\left( {a_3^2\left( {1308{\omega ^3} - 600\omega } \right) + 3{a_5}{a_3}\left( {2652{\omega ^3} - 920\omega } \right)} \right.} \\ {\left. { + 2\left( {5a_5^2\left( {1004{\omega ^3} - 248\omega } \right) + 4\left( {20{\omega ^3} - 18\omega } \right)} \right)} \right) + a_1^4\left( {3a_3^2\left( {636{\omega ^3} - 440\omega } \right) + 4{a_5}{a_3}\left( {1468{\omega ^3} - 760\omega } \right)} \right.} \\ {\left. { + 3\left( {a_5^2\left( {3324{\omega ^3} - 1080\omega } \right) + 4\left( {4{\omega ^3} - 6\omega } \right)} \right)} \right) + a_1^2\left( {3a_3^4\left( {2236{\omega ^3} - 760\omega } \right) + 32\omega \left( {{\omega ^2} - 1} \right)} \right.} \end{array} \hfill \\ \begin{array}{*{20}{l}} { + 18a_3^3{a_5}\left( {1364{\omega ^3} - 360\omega } \right) + 48a_5^2\left( {484{\omega ^3} - 78\omega } \right) + 9a_5^4\left( {5268{\omega ^3} - 680\omega } \right) + 4a_3^2\left( {4\left( {236{\omega ^3} - 90\omega } \right)} \right.} \\ { + 18a_3^3{a_5}\left( {1364{\omega ^3} - 360\omega } \right) + 48a_5^2\left( {484{\omega ^3} - 78\omega } \right) + 9a_5^4\left( {5268{\omega ^3} - 680\omega } \right) + 4a_3^2\left( {4\left( {236{\omega ^3} - 90\omega } \right)} \right.} \\ {\left. { - \tfrac{{1 - \pi }}{{8{\omega ^2}}}} \right)\, 5a_1^6{{\left( {{\omega ^2} - 2} \right)}^2} + a_1^5\left( {{\omega ^2} - 2} \right)\, \left( {5{a_3}\left( {11{\omega ^2} - 6} \right) + 3{a_5}\left( {9{\omega ^2} - 2} \right)} \right) + 5a_3^6{{\left( {2 - 9{\omega ^2}} \right)}^2}} \\ { + 3a_3^4\left( {a_5^2\left( {2911{\omega ^4} - 860{\omega ^2} + 60} \right) + 4\left( {81{\omega ^4} - 27{\omega ^2} + 2} \right)} \right) + 2{a_1}a_3^2{a_5}\left( {8\left( {287{\omega ^4} - 99{\omega ^2} + 6} \right)} \right.} \\ {\left. { + 2a_3^2\left( {1447{\omega ^4} - 620{\omega ^2} + 60} \right) + 15{a_5}{a_3}\left( {153{\omega ^4} - 52{\omega ^2} + 4} \right) + 3a_5^2\left( {3399{\omega ^4} - 940{\omega ^2} + 60} \right)} \right)} \\ { + a_5^2\left( {5a_5^4{{\left( {2 - 25{\omega ^2}} \right)}^2} + 12a_5^2\left( {625{\omega ^4} - 75{\omega ^2} + 2} \right) + 8{{\left( {1 - 25{\omega ^2}} \right)}^2}} \right) + a_3^2\left( {8{{\left( {1 - 9{\omega ^2}} \right)}^2}} \right. + } \\ {\left. {3a_5^4\left( {5631{\omega ^4} - 1180{\omega ^2} + 60} \right) + 16a_5^2\left( {803{\omega ^4} - 153{\omega ^2} + 6} \right)} \right) + 2a_1^3{a_3}\left( {a_3^2\left( {327{\omega ^4} - 300{\omega ^2} + 60} \right)} \right.} \end{array} \hfill \\ \begin{array}{*{20}{l}} {\left. { + 3{a_3}{a_5}\left( {663{\omega ^4} - 460{\omega ^2} + 60} \right) + 2\left( {5a_5^2\left( {251{\omega ^4} - 124{\omega ^2} + 12} \right) + 4\left( {5{\omega ^4} - 9{\omega ^2} + 2} \right)} \right)} \right)} \\ { + a_1^4\left( {3a_3^2\left( {159{\omega ^4} - 220{\omega ^2} + 60} \right) + 4{a_5}{a_3}\left( {367{\omega ^4} - 380{\omega ^2} + 60} \right)} \right. + 3\left( {4\left( {{\omega ^4} - 3{\omega ^2} + 2} \right)} \right.} \\ {\left. {\left. {a_5^2 + \left( {831{\omega ^4} - 540{\omega ^2} + 60} \right)} \right)} \right) + a_1^2\left( {3a_3^4\left( {559{\omega ^4} - 380{\omega ^2} + 60} \right) + 8{{\left( {{\omega ^2} - 1} \right)}^2}} \right.} \\ { + 9a_5^4\left( {1317{\omega ^4} - 340{\omega ^2} + 20} \right) + 48a_5^2\left( {121{\omega ^4} - 39{\omega ^2} + 2} \right) + 18a_3^3{a_5}\left( {341{\omega ^4} - 180{\omega ^2} + 20} \right)} \\ { + 4a_3^2\left( {3a_5^2\left( {1743{\omega ^4} - 700{\omega ^2} + 60} \right) + 4\left( {59{\omega ^4} - 45{\omega ^2} + 6} \right)} \right) + 6{a_3}{a_5}\left( {8\left( {49{\omega ^4} - 27{\omega ^2} + 2} \right)} \right.} \\ {\left. {\left. {\left. { + a_5^2 + \left( {2719{\omega ^4} - 860{\omega ^2} + 60} \right)} \right)} \right)} \right) = 0, } \end{array} \hfill \\ \end{gathered} \right)\,$

(34)

$\left. \begin{gathered} \begin{array}{*{20}{l}} {\tfrac{\pi }{{8\omega }}\, \left( {3a_1^5} \right.\left( {{\omega ^2} - 2} \right)\, \left( {9{\omega ^2} - 2} \right) + 6a_3^4{a_5}\left( {2911{\omega ^4} - 860{\omega ^2} + 60} \right) + 2a_1^3{a_3}\left( {3{a_3}\left( {663{\omega ^4} - 460{\omega ^2} + 60} \right)} \right.} \\ {\left. { + 20{a_5}\left( {251{\omega ^4} - 124{\omega ^2} + 12} \right)} \right) + a_1^4\left( {4{a_3}\left( {367{\omega ^4} - 380{\omega ^2} + 60} \right) + 6{a_5}\left( {831{\omega ^4} - 540{\omega ^2} + 60} \right)} \right)} \\ { + 2{a_1}a_3^2{a_5}\left( {15{a_3}\left( {153{\omega ^4} - 52{\omega ^2} + 4} \right) + 6{a_5}\left( {3399{\omega ^4} - 940{\omega ^2} + 60} \right)} \right) + 2{a_1}a_3^2\left( {8\left( {287{\omega ^4} - 99{\omega ^2} + 6} \right)} \right.} \\ {\left. { + 2a_3^2\left( {1447{\omega ^4} - 620{\omega ^2} + 60} \right) + 15{a_5}{a_3}\left( {153{\omega ^4} - 52{\omega ^2} + 4} \right) + 3a_5^2\left( {3399{\omega ^4} - 940{\omega ^2} + 60} \right)} \right)} \\ { + a_5^2\left( {20a_5^3{{\left( {2 - 25{\omega ^2}} \right)}^2} + 24{a_5}\left( {625{\omega ^4} - 75{\omega ^2} + 2} \right)} \right) + a_3^2\left( {32{a_5}\left( {803{\omega ^4} - 153{\omega ^2} + 6} \right)} \right.} \end{array} \hfill \\ \begin{array}{*{20}{l}} {\left. { + 12a_5^3\left( {5631{\omega ^4} - 1180{\omega ^2} + 60} \right)} \right) + 2{a_5}\left( {12a_5^2\left( {625{\omega ^4} - 75{\omega ^2} + 2} \right) + 8{{\left( {1 - 25{\omega ^2}} \right)}^2}} \right.} \\ {\left. { + 5a_5^4{{\left( {2 - 25{\omega ^2}} \right)}^2}} \right) + a_1^2\left( {18a_3^3\left( {341{\omega ^4} - 180{\omega ^2} + 20} \right) + 96{a_5}\left( {121{\omega ^4} - 39{\omega ^2} + 2} \right)} \right.} \\ { + 24{a_5}a_3^2\left( {1743{\omega ^4} - 700{\omega ^2} + 60} \right) + 12a_5^2{a_3}\left( {2719{\omega ^4} - 860{\omega ^2} + 60} \right)} \\ {\left. {\left. { + 36a_5^3\left( {1317{\omega ^4} - 340{\omega ^2} + 20} \right) + 6{a_3}\left( {a_5^2\left( {2719{\omega ^4} - 860{\omega ^2} + 60} \right) + 8\left( {49{\omega ^4} - 27{\omega ^2} + 2} \right)} \right)} \right)} \right) = 0, } \end{array} \hfill \\ \end{gathered} \right)\,$

(35)

$\left. \begin{gathered} \begin{array}{*{20}{l}} {\tfrac{1}{{8\omega }}\pi \left( {\, 5a_1^5} \right.\left( {{\omega ^2} - 2} \right)\, \left( {11{\omega ^2} - 6} \right) + 30a_3^5{{\left( {2 - 9{\omega ^2}} \right)}^2} + 2{a_1}a_3^2{a_5}\left( {4{a_3}\left( {1447{\omega ^4} - 620{\omega ^2} + 60} \right)} \right.} \\ {\left. { + 15{a_5}\left( {153{\omega ^4} - 52{\omega ^2} + 4} \right)} \right) + a_1^4\left( {6{a_3}\left( {159{\omega ^4} - 220{\omega ^2} + 60} \right) + 4{a_5}\left( {367{\omega ^4} - 380{\omega ^2} + 60} \right)} \right)} \\ { + 2a_1^3{a_3}\left( {2{a_3}\left( {327{\omega ^4} - 300{\omega ^2} + 60} \right) + 3{a_5}\left( {663{\omega ^4} - 460{\omega ^2} + 60} \right)} \right) + 12a_3^3\left( {4\left( {81{\omega ^4} - 27{\omega ^2} + 2} \right)} \right.} \\ {\left. { + a_5^2\left( {2911{\omega ^4} - 860{\omega ^2} + 60} \right)} \right) + 4{a_1}{a_3}{a_5}\left( {2a_3^2\left( {1447{\omega ^4} - 620{\omega ^2} + 60} \right) + 8\left( {287{\omega ^4} - 99{\omega ^2} + 6} \right)} \right.} \\ {\left. { + 3a_5^2\left( {3399{\omega ^4} - 940{\omega ^2} + 60} \right) + 15{a_3}{a_5}\left( {153{\omega ^4} - 52{\omega ^2} + 4} \right)} \right) + 2{a_3}\left( {8{{\left( {1 - 9{\omega ^2}} \right)}^2}} \right.} \end{array} \hfill \\ \begin{array}{*{20}{l}} {\left. { + 3a_5^4\left( {5631{\omega ^4} - 1180{\omega ^2} + 60} \right) + 16a_5^2\left( {803{\omega ^4} - 153{\omega ^2} + 6} \right)} \right) + 2a_1^3\left( {a_3^2\left( {327\omega _3^4 - 300{\omega ^2} + 60} \right)} \right.} \\ {\left. { + 3{a_3}{a_5}\left( {663{\omega ^4} - 460{\omega ^2} + 60} \right) + 2\left( {5a_5^2\left( {251{\omega ^4} - 124{\omega ^2} + 12} \right) + 4\left( {5{\omega ^4} - 9{\omega ^2} + 2} \right)} \right)} \right)} \\ { + a_1^2\left( {12a_3^3\left( {559{\omega ^4} - 380{\omega ^2} + 60} \right) + 54a_3^2{a_5}\left( {341{\omega ^4} - 180{\omega ^2} + 20} \right) + } \right.\, 8{a_3}\left( {4\left( {59{\omega ^4} - 45{\omega ^2} + 6} \right)} \right.} \\ {\left. {\left. {\left. { + 3a_5^2\left( {1743{\omega ^4} - 700{\omega ^2} + 60} \right)} \right) + 6{a_5}\left( {a_5^2\left( {2719{\omega ^4} - 860{\omega ^2} + 60} \right) + 8\left( {49{\omega ^4} - 27{\omega ^2} + 2} \right)} \right)} \right)} \right) = 0.} \end{array} \hfill \\ \end{gathered} \right)\,$

(36)

Putting ${a_1} = A - {a_3} - {a_5}$ in the minimization problem changes the constraint minimization problem to an unconstrained minimization problem, which is easier to find the solution of Eq (33) by using the conditions $\partial E({u_3})/\partial \omega = 0,$ $\partial E({u_3})/\partial {a_3} = 0$ and $\partial E({u_3})/\partial {a_5} = 0$ .

We plot the analytical solutions obtained from MGEMM (red line) Eq (32) and compare them with numerical solutions of Eq (23) obtained using the fourth order Runge-Kutta method (blue line). It is observed that for all different values of the amplitude A, the approximate solutions match extremely well with the numerical solutions (see Figure 2).

Figure 2. Comparison of the approximate solution (red line) with the numerical solution (blue line).

DownLoad: Full-Size Img PowerPoint

3.2.2. Case 2

Second, we consider ${k_1} = 0,$ ${k_2} = 1$ and ${k_3} = 1$ in Eq (6). Hence, we have the one-dimensional nonlinear oscillator governed by ^{[21,22,23,24,25,26]}

$\ddot u + \frac{{{u^3}}}{{1 + {u^2}}} = 0, \, \;\;\;\;\;u(0) = A, \, \;\;\;\dot u(0) = 0.$

(37)

Finally, we can obtain the first and second-order approximations to Eq (37) given by Eqs (25) and (28), respectively. We remark that the third-order approximation in Eq (32) can be given by MGEMM in a similar manner. Generally, after three steps of MGEMM, one can obtain the approximated solutions to Eq (37) with sufficient accuracy. The analytical results of Eq (37) are compared with the iterative homotopy harmonic balance method (IHHBM) ^[34], the energy balance method (EBM) ^[20], the max-min approach (MMA) ^[32], the global residue harmonic balance method (GRHBM) ^[29], the Hamiltonian approach (HA) ^[26] and the exact solutions, as shown in Table 3.

Table 3. Comparison of the approximate analytical frequencies with the exact solutions.

$A$	${\omega _{EBM}}$	${\omega _{IHHBM}}$	${\omega _{MMA}}$	${\omega _{GRHBM}}$	${\omega _{HA}}$	${\omega _{3rd\, GEMM}}$	${\omega _{exact}}$
	^[20]	^[34]	^[32]	^[29]	^[26]	present	Exact
0.01	0.00866	0.008478	0.00866	0.008472	0.00865	0.00847	0.00847
0.1	0.08627	0.084418	0.08627	0.084394	0.08624	0.08439	0.08439
1	0.65164	0.63136	0.65465	0.636795	0.64359	0.636783	0.63678
5	0.97343	0.96667	0.97435	0.968107	0.96731	0.969202	0.96698
10	0.99314	0.99090	0.99340	0.991591	0.99095	0.992005	0.99092
50	0.99973	0.99961	0.99973	0.999657	0.999608	0.999676	0.99961
100	0.99999	0.999901	0.99993	0.999914	0.99990	0.999919	0.99990

| Show Table

DownLoad: CSV

4. Results and discussion

In this paper, we test the analytical solutions of strongly nonlinear Duffing-harmonic oscillators to show the effectiveness of MGEMM. Comparisons of the analytical solutions and the exact numerical solutions of the Duffing-harmonic oscillators for small and large values of the amplitude have been illustrated in Figures 1 and 2 and Tables 1–3. In these Figures, the analytical solution is indicated by a red line, while the numerical solution is represented by a blue line. There is good compatibility between the analytical and numerical solutions, which confirms the accuracy of our results, and excellent matching is observed in these calculations. Moreover, as shown in Tables 1–3, the results produced using the MGEMM are in better agreement with those obtained using exact solutions than other existing ones in the literature. The calculations of the above applications were done using Mathematica software.

5. Conclusions

In this paper, the modified global error minimization method has been presented successfully to obtain higher order approximate periodic solutions of strongly nonlinear Duffing-harmonic oscillators. The present method, which is proved to be a powerful mathematical tool to study nonlinear oscillators, can be easily extended to any nonlinear equation because of its efficiency and convenient applicability. We demonstrated the accuracy and efficiency of the proposed method by solving some examples. We showed that the obtained solutions are valid for the whole domain. Comparisons of the obtained solutions, whether numerical or analytical, revealed a clear match, highlighting the precision of the modified GEMM.

Acknowledgments

M. Zayed and G. M. Ismail extend their appreciation to the Deanship of Scientific Research at King Khalid University, Saudi Arabia, for funding this work through the research groups program under grant R.G.P.2/207/43.

Conflict of interest

The authors declare that they have no conflict of interest.

References

[1]	S. Coleri, M. Ergen, A. Puri, A. Bahai, Channel estimation techniques based on pilot arrangement in OFDM systems, IEEE Trans. Broadcast., 48 (2002), 223–229. https://doi.org/10.1109/TBC.2002.804034 doi: 10.1109/TBC.2002.804034
[2]	Y. Gong, K. B. Letaief, Low rank channel estimation for space-time coded wideband OFDM systems, in IEEE 54th Vehicular Technology Conference. VTC Fall 2001. Proceedings (Cat. No.01CH37211), 2 (2001), 772–776. https://doi.org/10.1109/VTC.2001.956875
[3]	M. Morelli, U. Mengali, A comparison of pilot-aided channel estimation methods for OFDM systems, IEEE Trans. Signal Process., 49 (2001), 3065–3073. https://doi.org/10.1109/78.969514 doi: 10.1109/78.969514
[4]	Y. Li, L. J. Cimini, N. R. Sollenberger, Robust channel estimation for OFDM systems with rapid dispersive fading channels, IEEE Trans. Commun., 46 (1998), 902–915. https://doi.org/10.1109/26.701317 doi: 10.1109/26.701317
[5]	Y. S. Cho, J. Kim, W. Y. Yang, C. G. Kang, MIMO-OFDM Wireless Communications with MATLAB, Wiley Publishing, 2010. https://doi.org/10.1002/9780470825631.ch4
[6]	J. Long, Study and simulation on channel estimate algorithm in OFDM system, Commun. Technol., 41 (2008), 7–8. https://doi.org/10.3969/j.issn.1002-0802.2008.10.003 doi: 10.3969/j.issn.1002-0802.2008.10.003
[7]	D. Wang, Z. Mei, J. Liang, J. Liu, An improved channel estimation algorithm based on WD-DDA in OFDM system, Mobile Inf. Syst., 2021 (2021), 6540923. https://doi.org/10.1155/2021/6540923 doi: 10.1155/2021/6540923
[8]	Y. Li, C. Tao, G. Secogranados, A. Mezghani, A. L. Swindlehurst, L. Liu, Channel estimation and performance analysis of One-Bit massive mimo systems, IEEE Trans. Signal Process., 65 (2017), 4075–4089. https://doi.org/10.1109/TSP.2017.2706179 doi: 10.1109/TSP.2017.2706179
[9]	H. He, C. K. Wen, J. Shi, G. Y. Li, Deep learning-based channel estimation for beamspace mmwave massive mimo systems, IEEE Wireless Commun. Lett., 7 (2018), 852–855. https://doi.org/10.1109/LWC.2018.2832128 doi: 10.1109/LWC.2018.2832128
[10]	M. Soltani, V. Pourahmadi, A. Mirzaei, H. Sheikhzadeh, Deep learning-based channel estimation, IEEE Commun. Lett., 23 (2019), 652–655. https://doi.org/10.1109/LCOMM.2019.2898944 doi: 10.1109/LCOMM.2019.2898944
[11]	H. Ye, G. Y. Li, B. H. Juang, Power of deep learning for channel estimation and signal detection in OFDM systems, IEEE Wireless Commun. Lett., 7 (2017), 114–117. https://doi.org/10.1109/LWC.2017.2757490 doi: 10.1109/LWC.2017.2757490
[12]	A. S. M. Mohammed, A. I. A. Taman, A. M. Hassan, A. Zekry, Deep learning channel estimation for OFDM 5G systems with different channel models, Wireless Pers. Commun., 128 (2023), 2891–2912. https://doi.org/10.1007/s11277-022-10077-6 doi: 10.1007/s11277-022-10077-6
[13]	S. A. Fechtel, A novel approach to modeling and efficient simulation of frequency-selective fading radio channels, IEEE J. Sel. Areas Commun., 11 (1993), 422–431. https://doi.org/10.1109/49.219555 doi: 10.1109/49.219555
[14]	S. Ioffe, C. Szegedy, Batch normalization: accelerating deep network training by reducing internal convariate shift, in Proceedings of the 32nd International Conference on Machine Learning, 37 (2015), 448–456. Available from: http://proceedings.mlr.press/v37/ioffe15.html.

This article has been cited by:

1.	Nadia M. Farea, Gamal M. Ismail, Analytical solution for strongly nonlinear vibration of a stringer shell, 2024, 1461-3484, 10.1177/14613484241278421
2.	Gamal M. Ismail, Alwaleed Kamel, Abdulaziz Alsarrani, Approximate analytical solutions to nonlinear oscillations via semi-analytical method, 2024, 98, 11100168, 97, 10.1016/j.aej.2024.04.040
3.	Livija Cveticanin, Miodrag Zukovic, Dragan Cveticanin, Approximate Analytic Frequency of Strong Nonlinear Oscillator, 2024, 12, 2227-7390, 3040, 10.3390/math12193040
4.	Nazmul Sharif, Helal Uddin Molla, M.S. Alam, Analytical solution of couple-mass-spring systems by novel homotopy perturbation method, 2024, 167, 00207462, 104923, 10.1016/j.ijnonlinmec.2024.104923
5.	Gamal M. Ismail, Nadia M. Farea, Mahmoud Bayat, Ji Wang, Investigation of the highly complex nonlinear problems via modified energy balance method, 2024, 23071877, 10.1016/j.jer.2024.07.006
6.	Gamal M. Ismail, Alwaleed Kamel, Abdulaziz Alsarrani, Studying nonlinear vibration analysis of nanoelectro-mechanical resonators via analytical computational method, 2024, 22, 2391-5471, 10.1515/phys-2024-0011
7.	Zhenbo Li, Jin Cai, Linxia Hou, A modified generalized harmonic function perturbation method and its application in analyzing generalized Duffing–Harmonic–Rayleigh–Liénard oscillator, 2024, 166, 00207462, 104832, 10.1016/j.ijnonlinmec.2024.104832

Reader Comments

Your name:*

Email:*
© 2023 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)

通讯作者: 陈斌, bchen63@163.com

1.
沈阳化工大学材料科学与工程学院沈阳 110142

Electronic Research Archive

1 1.3

Metrics

Article views(2183) PDF downloads(100) Cited by(2)

Preview PDF

Download XML

Export Citation

Article outline

Show full outline

Electronic Research Archive

An improved least squares (LS) channel estimation method based on CNN for OFDM systems

Related Papers:

Abstract

1. Introduction

2. Basic idea of the modified global error minimization method (MGEMM)

3. Examples of two nonlinear Duffing-harmonic oscillators

3.1. Example 1

3.1.1. Case 1

3.1.2. Case 2

3.2. Example 2

3.2.1. Case 1

3.2.2. Case 2

4. Results and discussion

5. Conclusions

Acknowledgments

Conflict of interest

References

This article has been cited by:

Reader Comments

通讯作者: 陈斌, bchen63@163.com

Metrics

Figures and Tables

Other Articles By Authors

Catalog

Abstract

1. Introduction

2. Basic idea of the modified global error minimization method (MGEMM)

3. Examples of two nonlinear Duffing-harmonic oscillators

3.1. Example 1

3.1.1. Case 1

3.1.2. Case 2

3.2. Example 2

3.2.1. Case 1

3.2.2. Case 2

4. Results and discussion

5. Conclusions

Acknowledgments

Conflict of interest

References

Electronic Research Archive

An improved least squares (LS) channel estimation method based on CNN for OFDM systems

Related Papers:

Abstract

1. Introduction

2. Basic idea of the modified global error minimization method (MGEMM)

3. Examples of two nonlinear Duffing-harmonic oscillators

3.1. Example 1

3.1.1. Case 1

3.1.2. Case 2

3.2. Example 2

3.2.1. Case 1

3.2.2. Case 2

4. Results and discussion

5. Conclusions

Acknowledgments

Conflict of interest

References

This article has been cited by:

Reader Comments

通讯作者: 陈斌, bchen63@163.com

Metrics

Figures and Tables

Other Articles By Authors

Related pages

Tools

Export File

Citation

Format

Content

Catalog

Abstract

1. Introduction

2. Basic idea of the modified global error minimization method (MGEMM)

3. Examples of two nonlinear Duffing-harmonic oscillators

3.1. Example 1

3.1.1. Case 1

3.1.2. Case 2

3.2. Example 2

3.2.1. Case 1

3.2.2. Case 2

4. Results and discussion

5. Conclusions

Acknowledgments

Conflict of interest

References