Virtual Axle Detector Based on Analysis of Bridge Acceleration Measurements by Fully Convolutional Network

Area of Science:

• Civil Engineering
• Structural Health Monitoring
• Signal Processing

Background:

• Accurate axle detection is crucial for Bridge Weigh-In-Motion (BWIM) systems.
• Conventional methods often require specific bridge designs or dedicated sensors.
• Arbitrary accelerometer placement offers a flexible alternative for data acquisition.

Purpose of the Study:

• To develop a novel, bridge-type-independent method for axle detection using accelerometers.
• To implement a simplified binary classification model for robust axle detection.
• To establish Virtual Axle Detectors (VADs) without traditional axle detectors.

Main Methods:

• Utilized accelerometers placed arbitrarily on a bridge structure.
• Developed a Fully Convolutional Network (FCN) model for signal processing.
• Processed acceleration signals as Continuous Wavelet Transforms (CWTs) for multi-scale analysis.

Main Results:

• Achieved 95% axle detection accuracy on a steel trough railway bridge.
• Correctly detected 128,599 out of 134,800 previously unseen axles.
• 90% of axles were detected with a maximum spatial error of 20 cm at speeds up to 56.3 m/s.

Conclusions:

• The proposed method effectively uses accelerometers as Virtual Axle Detectors (VADs).
• The FCN model demonstrates high accuracy and efficiency in real-world operating conditions.
• This approach offers a versatile solution for axle detection in BWIM applications.