Usage of Machine Learning Techniques to Classify and Predict the Performance of Force Sensing Resistors

Area of Science:

• Materials Science
• Electrical Engineering
• Computational Science

Background:

• Advancements in polymer-based sensor manufacturing aim to enhance properties like sensitivity and reduce drift and hysteresis.
• Novel manufacturing methods include additive manufacturing, microfluidic preparation, and brush painting.
• Significant potential exists to improve sensor performance through computational techniques like modeling and machine learning.

Purpose of the Study:

• To investigate the potential of modeling, classification, and machine learning to enhance polymer-based sensor performance.
• To offer inexpensive computational methods as an alternative to complex manufacturing routes for end-users.
• To characterize Force Sensing Resistors (FSRs) for drift and hysteresis errors.

Main Methods:

• Characterization of 96 FSR specimens (two commercial brands) under varying input voltages.
• Electrical measurements of output voltage at null force (V) to correlate with sensor errors.
• Application of k-means clustering for sensor classification based on V readings.
• Presentation of theoretical foundations of FSRs and their correlation with modeling/classification techniques.

Main Results:

• A significant inverse correlation was found between the output voltage at null force (V) and drift error.
• Sensor performance can be predicted through inexpensive electrical measurements before deployment.
• No direct relationship was found between V and hysteresis error.
• K-means clustering successfully distinguished between high and low hysteresis sensors using only V readings.

Conclusions:

• Inexpensive electrical measurements can predict drift error in FSRs.
• Machine learning classification offers a viable method to pre-emptively identify sensors with high hysteresis.
• Computational approaches provide a cost-effective means to improve and select polymer-based sensors.