Multi-sensor integration for on-line tool wear estimation through radial basis function networks and fuzzy neural...

Area of Science:

• Manufacturing Engineering
• Artificial Intelligence in Manufacturing
• Sensors and Signal Processing

Background:

• Online tool wear estimation is crucial for industry automation, impacting productivity and product quality.
• Timely tool replacement decisions are essential in machining systems to maintain efficiency.
• Existing methods may lack the accuracy and adaptability required for dynamic machining environments.

Purpose of the Study:

• To develop an advanced online tool wear estimation system by integrating artificial neural networks (ANN) and fuzzy logic.
• To improve the accuracy and reliability of tool wear prediction in real-time machining processes.
• To enhance decision-making for tool changes, optimizing machining operations.

Main Methods:

• Proposed a five-component online estimation system: data collection, feature extraction, pattern recognition, multi-sensor integration, and tool/work distance compensation.
• Employed radial basis function (RBF) networks for feature recognition from individual sensors.
• Developed a fuzzy neural network (FNN) model for integrating multi-sensor decisions, featuring self-organization and self-adjustment capabilities.

Main Results:

• The proposed system demonstrated significant improvements in the accuracy of product profile.
• The fuzzy neural network model effectively integrated information from multiple sensors for robust estimation.
• Experimental validation on metal cutting processes confirmed the system's performance.

Conclusions:

• The integrated ANN and fuzzy logic system provides an effective solution for online tool wear estimation.
• The developed system enhances machining accuracy and supports timely tool management.
• This approach offers a self-learning and adaptive method for improving industrial automation.