ESTIMATING THE LIFESPAN OF AIR PURIFIER FILTERS IN SPORTS FACILITIES USING NEURAL NETWORK MODELS

Abstract

Objective: This study aims to develop and optimize a neural network-based method for predicting the lifespan of air purifier filter elements in sports facilities, enhancing air quality management and ensuring healthier training environments. Methods: Usage data from air purifiers and characteristics of filter elements were collected across various sports facilities. A multi-layer perceptron (MLP) neural network model was initially established to predict the lifespan of these filters. To refine accuracy, a Genetic Algorithm (GA) was employed to optimize the MLP model. The study involved comparing the performance of two different types of air purifier filter cartridges, analyzing data to assess the predictive accuracy of the MLP model before and after GA optimization. Results: The MLP model initially showed a maximum absolute value of the relative error in lifespan prediction for the two filter types at 15.6% and 10.98%. After optimization with the Genetic Algorithm, this error was significantly reduced to 4.64% and 4.88%, respectively. These improvements demonstrate the effectiveness of the GA-optimized MLP model in providing more precise lifespan predictions. Conclusion: The GA-optimized MLP neural network model effectively predicts the lifespan of filter elements in air purifiers used within sports facilities. This enhanced predictive capability allows for better management of air quality and filter maintenance, contributing to safer and more effective athletic training environments. This study not only supports facility managers in maintaining high air quality standards but also serves as a benchmark for further research in applying advanced predictive technologies in sports facility management.