AI Filtering System for Optimizing Wavelet-Based Industrial Communications

This is the authors' abstract. We don't add key points for this paper.

Wireless communication systems in industrial environments face many challenges, including interference from noisy environments, complex wireless resource management and multipath issues affecting transmission reliability. To address these issues, we suggest architectures based on the discrete wavelet packet transform, a crucial process for separating noise and examining non-stationary signals. In harsh industrial environments, where latency is extremely low, reliability high and data rates considerable, this method is of paramount importance. These architectures offer exceptional performance, even in the presence of complex conditions such as multiple noise sources, generally considered as impulse noise. This paper presents an innovative method based on the use of artificial intelligence (AI) for noise filtering in reception. We have examined the effectiveness of various AI-based filtering models, such as Artificial Neural Networks (ANN), Random Forest (RF) and Support Vector Machines (SVM). The accuracy of these models was compared, highlighting their ability to significantly optimize the quality of industrial communications. ANNs offered the best performance, with an RMSE of 0.0228, although this increased accuracy was accompanied by higher complexity and training time. RF followed with an RMSE of 0.165 and a coefficient of determination ( R 2 ) of 0.94, offering better interpretability and notable speed. The SVM model, with an RMSE of 0.23262 and R 2 of 0.88, showed good explanatory power despite lower accuracy than the other models.