Research on inherent nonlinearity calibration of micro displacement sensor using neural network

DOI:	10.11809/bqzbgcxb2025.01.023
Keywords:	displacement sensor; nonlinear calibration model; neural network method; measurement accuracy; real time computation
Abstract:	Micro carbon film displacement sensor has a broad application background in the fields of agricultural machinery, robot end effectors, medical and surgical instruments due to its advantages of compact structure, stability, low cost and so on. Due to the manufacturing error of carbon film thickness, the micro carbon film displacement sensor has an inherent nonlinearity, which could greatly affect its measurement accuracy. To calibrate the inherent nonlinearity of the micro displacement sensor, a Neural Network based Calibration Method (NNCM) is presented. By employing a methodology that combines simulation with experimentation, a comparative study was conducted to evaluate the NNCM against existing PCM and BCM across two dimensions, those are correction accuracy and real time computing speed. Results reveal that increasing the model order can effectively enhance the calibration accuracy. For the BCM and NNCM, convergence in accuracy was achieved with the third order model. Through the third order PCM, BCM, and NNCM calibration, the measurement errors of the sensor were reduced by 46.1%, 89.0%, and 89.6% respectively. Therefore, the NNCM demonstrated the better precision in nonlinearity calibration. The real time computing costs of PCM, BCM, and NNCM are 0.48 ms, 0.49 ms, and 0.85 ms, respectively, which could generally meet the requirements for applications with high performance controllers at the 5 ms level.
Issue:	Vol. 46 No. 1 (2025)
Published:	2025-01-31
PDF