Artificial Neural Network-Based Conveying Object Measurement Automation System Using Distance Sensor

Area of Science:

• Industrial Engineering
• Automation and Control Systems
• Artificial Intelligence in Logistics

Background:

• Measurement technology is crucial for logistics operations like defect inspection and loading optimization.
• The fourth industrial revolution drives research into measurement automation using AI, IoT, and advanced sensors.
• High costs of 3D scanners and reliability issues with entry-level sensors hinder widespread adoption in logistics.

Purpose of the Study:

• To develop a systematic framework for reliable geometry measurement using a single, low-cost distance sensor.
• To enable accurate estimation of object length and width in logistics environments.
• To bridge the gap between high-performance, expensive measurement systems and unreliable entry-level sensors.

Main Methods:

• Designed and constructed a conveyor-belt data acquisition setup simulating realistic logistics transfer scenarios.
• Collected measurement data under systematically varied transfer conditions to capture environmental disturbances.
• Employed robust feature extraction for noisy, condition-dependent signals and trained an artificial neural network (ANN) for dimension estimation.

Main Results:

• The proposed framework successfully maps sensor observations to geometric dimensions (length and width).
• The ANN model demonstrated reliable performance in estimating object dimensions using test data.
• Experimental results confirm the method's robustness even under diverse and challenging transfer conditions.

Conclusions:

• A cost-effective and reliable solution for geometry measurement in logistics has been demonstrated.
• The proposed framework significantly improves the usability of entry-level distance sensors for industrial applications.
• This approach offers a practical alternative to expensive measurement technologies in the logistics sector.