.png){kind=icon}
.png){kind=icon}
Return
A two-stage linearized model for thermistor circuit linearization using LSTM based multi-layer self-attention model
.png){kind=icon}
PRE
AI
DOI：10.1016/j.measurement.2025.116733.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In industrial processes, precise temperature measurement is crucial, with thermistors being favored for their inherent sensitivity, compact size, and rapid response times. However, the nonlinear characteristics of Negative Temperature Coefficient (NTC) thermistors pose challenges in achieving linearity across their operational range. To address this, a novel two-stage linearization technique combining analog voltage controlled oscillator- signal conditioning circuit (VCO-SCC) and digital long short-term memory- multi-layer self-attention (LSTM-MLSA) methods is proposed. The VCO-SCC partially linearizes thermistor characteristics, with frequency-determining parameters tailored to the thermistor's parameters and operating range. Subsequently, the LSTM-MLSA model further refines nonlinearities in the VCO-SCC output, enhancing overall linearity. Experimental validation using three NTC thermistors demonstrates that both methods yield frequency outputs closely resembling the desired values, with LSTM-MLSA showing smaller deviations and errors compared to VCO-SCC. Regression analysis confirms LSTM-MLSA's superior performance across all thermistors, exhibiting lower RSS, higher Adjusted RSquare, and smaller errors. Therefore, the proposed technique offers improved accuracy and precision in predicting frequency output, making LSTM-MLSA the preferable choice for temperature sensing applications across diverse thermistors. The study contributes to the development of robust linearization methods for NTC thermistors, crucial for ensuring optimal performance and reliability in industrial temperature sensing.
Keywords:
Temperature measurement
Thermistors
Linearization technique
Analog and digital methods
LSTM-MLSA
Journal
IF:
5.6
Papers:
2.0W
Citations:
5.4W
View more organizations