Innovative rapid liquid concentration measurement based on thermal lens effect and machine learning

Area of Science:

• Optics and Photonics
• Analytical Chemistry
• Machine Learning Applications

Background:

• The thermal lens effect (TLE) is well-studied for laser system analysis but underutilized for solution concentration quantification.
• Accurate and rapid online measurement of liquid concentrations is crucial for industrial process control and quality assurance.

Purpose of the Study:

• To explore the relationship between liquid concentrations and TLE-induced interference fringes.
• To develop and validate a novel TLE-based method for real-time solution concentration measurement.
• To integrate image processing, DFT, and machine learning for enhanced measurement accuracy.

Main Methods:

• Utilizing the thermal lens effect (TLE) to induce interference fringes in solutions of varying concentrations.
• Applying image processing and discrete Fourier transform (DFT) for extracting features from interference ring patterns.
• Employing a backpropagation artificial neural network (BP-ANN) for modeling and predicting solution concentrations.

Main Results:

• Demonstrated a high accuracy in concentration measurement with low root mean square error (RMSE) values.
• Achieved RMSE of 3.055 for the training set and 5.396 for the test set, validating the model's predictive power.
• Successfully enabled precise, real-time determination of soy sauce concentration.

Conclusions:

• The developed TLE-based method, enhanced by image processing, DFT, and BP-ANN, offers a viable solution for rapid, online liquid concentration measurement.
• This approach has significant potential for industrial testing, quality control, environmental monitoring, and other fields requiring precise concentration analysis.
• The study highlights the underappreciated utility of TLE in analytical chemistry and process monitoring.