Sensor Optimization and Regression Analysis for Fan Performance Prediction Using Vibration Data

This study identifies representative sensors for monitoring fan performance by analyzingvibration data collected from piezoelectric sensors during various operational modes. Thedataset, which includes measurements at a rate of 300 samples/sec from 10 sensors,covers six modes of operation: Maximum Speed, Maximum Speed with Oscillation,Minimum Speed, Minimum Speed with Oscillation, Minimum to Maximum Speed, and acomprehensive dataset combining all modes. Using statistical methods such ascorrelation matrices and eigenvalues, we identify sensors with strong correlations to thefan’s behaviour. Regression techniques are applied to predict the target variable, and theMean Squared Error (MSE) is used to assess the accuracy of the models. This analysisfacilitates sensor optimization, ensuring the minimum number of sensors required foraccurate predictions.