International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 11 | Nov 2025
p-ISSN: 2395-0072
www.irjet.net
Neural Networks for Modelling Nonlinear Oscillations in Mechanical Systems Haran S S1, Harish S S2, Shanthi Chinnasamy3 1Student, Department of Artificial Intelligence and Data Science, RVS College of Engineering and Technology,
Coimbatore, Tamil Nadu, India
2Former Student, Department of Physics, Bharathiar University, Coimbatore, Tamil Nadu, India 3Independent Author, Coimbatore, Tamil Nadu, India
-----------------------------------------------------------------------***-----------------------------------------------------------------------ABSTRACT Nonlinear oscillations are intrinsic to a wide range of mechanical systems. Modelling the system behaviour is crucial for stability, error identification and optimization of the system parameters. Applying traditional analytical and mathematical expressions to the system often finds difficult due to the nonlinear oscillations. Thus, the advancement of Artificial Intelligence in mechanical systems have transformed this traditional approach into neural network architectures including feedforward, recurrent, convolutional, and physics-informed networks in capturing nonlinear behaviours in mechanical oscillators. It discusses the IoT embedded systems and data acquisition for real-time monitoring of nonlinear oscillations. It also highlights the data processing and feature extraction methods to improve the systemโs performance. Furthermore, Evaluation and validation of errors in a model is also discussed by statistical approaches. Finally, this review explores emerging trends in neural networks such as hybrid AI-physics model and their integration within the industry and smart manufacturing ecosystem. Key Words: Artificial Neural Network, Nonlinear Oscillations, Mechanical Systems, IoT systems, Data Processing, Engineering Models, Model Validation
1. INTRODUCTION Nonlinear oscillations are fundamental phenomena that creates the foundation of many fields in technical sciences from micro-scale resonators and vibration absorbers to large-scale applications. The main goal is to investigate the level of vibration in the mechanical systems. Unlike linear system, nonlinear systems exhibit complex behaviour such as amplitude-dependent frequencies, bifurcations, chaos, and energy transfer between modes. Traditional methods like analytical and mathematical methods like Perturbation, phase-plane, bifurcation, averaging, runge-kutta methods often face limitations while dealing with nonlinear dynamics. This issue can be tackled by modelling a nonlinear response that is crucial for the systemโs stability and behaviour. This review paper provides a comprehensive overview of recent trends in applying neural networks for a nonlinear oscillation in mechanical system. Advanced architectures like Recurrent Neural Networks (RNNs), Long Short-Term Memory (LSTM) networks, and Physics-Informed Neural Networks (PINNs) are used to optimize the system behaviour. By bridging a gap between Nonlinear dynamics and Artificial Intelligence, this paper aims to provide insights into how neural networks can predict and control the nonlinear oscillations in mechanical systems.
2. OVERVIEW OF NONLINEAR OSCILLATIONS IN MECHANICAL SYSTEM Oscillations are the periodic variation of some measure around an equilibrium point between two or more different states. In linear oscillations, the restoring force is proportional to displacement. Hence, the behaviour of the system is clearly expressed by linear differential equations. ๐น = โ๐๐ฅ But in actual occurrences, the mechanical system doesnโt stick to such linearity. The oscillatory motion is said to be nonlinear when the restoring force is not directly proportional to the displacement. Nonlinear oscillations are essential to study the dynamics of mechanical systems as their behaviour is much complex than the linear ones. This nonlinearity arises due to nonlinear damping, large amplitude motions or complex part interactions. It means that systemโs output is not directly proportional to its input. Using the nonlinear differential equations, can predict the multiple stable states,
ยฉ 2025, IRJET
|
Impact Factor value: 8.315
|
ISO 9001:2008 Certified Journal
|
Page 215