Abstract:The dynamic calibration accuracy of pressure sensors with shock tube systems is often limited by the distorted calibration signal due to the interferences of nonuniform shock waves, vibrations, and complex noise. Signal decomposition has been verified to be a feasible way to correct the distorted calibration signals. However, the component loss problem in signal decompositions is a pain point that reduces the reliability of the calibration signal. In this article, a convolutional neural network (CNN)-assisted adaptive signal decomposition method is proposed to solve the component loss problem to further improve the calibration signal quality. First, the empirical mode decomposition (EMD) is used to adaptively extract the coarse ringing component from the distorted calibration signal. To effectively compensate the component loss problem of the coarse ringing component, a loss compensation model (LCM) is established by the CNN. Meantime, a dataset construction algorithm is proposed to improve the LCM accuracy based on the amplitude adjustment range and the autocorrelation function (ACF). The corrected calibration signal is finally obtained by the sum of the reconstructed ringing component and the preestimated trend from EMD results. A series of simulations and actual calibration experiments for two types of pressure sensors are carried out with a shock tube system to verify the performance of the proposed method. Results show that the proposed method is able to solve the component loss problem in signal decompositions and improve the quality of the calibration signals effectively. The mean absolute error (MAE) and root mean squared error (RMSE) of the amplitude-frequency characteristic (AFC) and phase-frequency characteristic (PFC) obtained by the proposed method are 0.13 and 0.31 V as well as 5.57° and 8.49°, respectively, which is about 1/8 and 1/5 as well as 1/6 and 1/6 times of the values obtained by the distorted signal. Furthermore, comparative experiments also demonstrate the superiority of the proposed method over the frequently-used methods in both calibration signal correction capability and robustness.

Calibration Of The Pressure Sensor Device With The Extreme Learning Machine

Calibrating Low-Cost Sensors by a Two-Phase Learning Approach for Urban Air Quality Measurement

Observability-Aware Intrinsic and Extrinsic Calibration of LiDAR-IMU Systems

A Sensorless Adaptive Optics Control System for Microscopy Based on Extreme Learning Machine

Research On Soft Sensor Based On Extreme Learning Machine

Application of Extreme Learning Machine to Gas Flow Measurement with Multipath Acoustic Transducers

Integrating Extreme Learning Machine with Kalman Filter to Bridge GPS Outages

Nonlinear industrial soft sensor development based on semi-supervised probabilistic mixture of extreme learning machines

Temperature Compensation and Pressure Bias Estimation for Piezoresistive Pressure Sensor Based on Machine Learning Approach

Deep Learning of Semisupervised Process Data with Hierarchical Extreme Learning Machine and Soft Sensor Application

Evaluations on supervised learning methods in the calibration of seven-hole pressure probes

Low-cost Measurement of Industrial Shock Signals via Deep Learning Calibration

Machine-Learning-Based Calibration of Temperature Sensors

Accurate Nonlinearity and Temperature Compensation Method for Piezoresistive Pressure Sensors Based on Data Generation.

Research on Temperature Compensation of Multi-Channel Pressure Scanner Based on an Improved Cuckoo Search Optimizing a BP Neural Network

An operational calibration approach of industrial robots through a motion capture system and an artificial neural network ELM

CNN-Assisted Adaptive Signal Decomposition Method for Correcting the Distorted Calibration Signals of Pressure Sensors

Comparison and Verification of Three Algorithms for Accuracy Improvement of Quartz Resonant Pressure Sensors.

A nonlinear regression model of pressure sensor based on SVM

Non-invasive cuff-less blood pressure estimation using a hybrid deep learning model

Continual Learning for Cuffless Blood Pressure Measurement Using PPG and ECG Signals