A neural network-evolutionary computational framework for remaining useful life estimation of mechanical systems

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Predictive Maintenance

Background:

• Accurate estimation of remaining useful life (RUL) is crucial for proactive maintenance of mechanical systems.
• Existing methods often require complex models, limiting their deployment in resource-constrained environments.

Purpose of the Study:

• To develop a novel framework for estimating the RUL of mechanical systems.
• To enable the deployment of RUL estimation in embedded systems with limited computational resources.

Main Methods:

• A framework combining a multi-layer perceptron and an evolutionary algorithm for parameter optimization.
• Utilizes a strided time window and a piecewise linear model for RUL estimation.
• Parameter tuning facilitates the use of compact neural network models.

Main Results:

• The proposed method demonstrates superior performance compared to state-of-the-art techniques on the C-MAPSS dataset.
• Achieves high accuracy in RUL estimation with a computationally efficient and compact model.
• Successfully evaluated on a publicly available benchmark dataset.

Conclusions:

• The developed framework offers an effective and efficient approach for RUL estimation in mechanical systems.
• The compact model design makes it suitable for deployment on embedded systems.
• This research advances predictive maintenance capabilities through optimized AI models.