Comparison of signal processing methods considering their optimal parameters using synthetic signals in a heat exchanger network simulation

Plant sensor data contain errors that can hamper process analysis and decision-making. Those dataset are not used to their full potential due to the complexity of their processing. This paper addresses these challenges by comparing popular data processing techniques based on their ability to process sensor data, all that while using optimal parameters. The latter are obtained for all approaches using an algorithm that performs a parametric sweep. The performance of Kalman filter, exponential weighted moving average filter, short-time Fourier transform, and wavelet transform to process synthetic flow and temperature signals from a heat exchanger network simulation is quantified given two criteria: signal-to-noise ratio (SNR) and root mean square error (RMSE). It is found that most of the time, wavelet transform showed the highest RMSE reduction and SNR improvement; the wavelet transform can effectively filter signals from distinct variables from a heat exchanger network simulation when optimal parameters are selected.