Speckle-decoded temperature-insensitive strain identification in a multimode optical fiber.

Area of Science:

• Optoelectronics
• Fiber-optic sensing
• Machine learning

Background:

• Fiber-optic sensors are vital for measuring physical and biochemical parameters.
• Temperature cross-sensitivity is a major limitation in practical fiber-optic sensing applications.
• Accurate strain measurement is crucial in various engineering fields.

Purpose of the Study:

• To develop a temperature-insensitive fiber-optic strain sensor.
• To utilize deep learning for enhanced signal recognition in optical sensors.
• To address the challenge of temperature cross-sensitivity in strain sensing.

Main Methods:

• A speckle-decoded sensing mechanism based on scattering patterns was employed.
• A deep learning algorithm was leveraged to analyze speckle patterns for strain estimation.
• Experimental validation was performed to assess sensor performance under varying temperatures and strains.

Main Results:

• The proposed sensor demonstrated temperature-insensitive strain measurement.
• A classification model achieved 99.28% recognition accuracy for axial strain (0-0.3 N).
• Strain prediction yielded an average root-mean-square error of 1.02 N%.

Conclusions:

• The intelligent speckle sensing strategy effectively mitigates temperature cross-sensitivity.
• Deep learning-based analysis of speckle patterns enables accurate and reliable strain sensing.
• This approach holds significant potential for advancing fiber-optic sensor applications in engineering.