Predictive Modeling of LED Luminaire Degradation Using Artificial Neural Networks and Environmental Input

One of the key requirements in the design of lighting systems is the long-term reliability of LED luminaires. This aspect is particularly emphasized in smart buildings and infrastructure, where maintenance must be optimized both economically and functionally. Despite the excellent properties of LEDs in terms of energy efficiency and lifetime, their components inevitably degrade over time due to thermal, electrical and environmental stresses. This leads to a gradual decrease in luminous and photometric performance. This paper proposes a hybrid predictive framework to estimate the degradation of LEDs by calculating a wear index based solely on indirect inputs such as daylight intensity, daily energy consumption and operating time. In this work, two multilayer artificial neural network models were developed: one to predict daily energy consumption, the other to estimate the average operating time of luminaires. The outputs of these models are then used to calculate the wear index using a derived formula based on the principles of physical degradation. Experimental validation using real measurement data shows that the proposed models achieve high predictive accuracy with relative errors of around 5% for partial parameters and below 10% for the final estimate of the wear index. The results presented in the paper confirm the suitability of models based on artificial neural networks for predicting the degradation of LED diodes without the need for direct operational monitoring or measurement of operating time. The proposed approach enables the implementation of cost-effective predictive maintenance strategies, which allows for the early identification of luminaires approaching the end of their functional life. This method supports the concept of proactive lighting management in smart environments, ensuring visual comfort, operational continuity and extended system sustainability. This approach eliminates the need for direct luminaire runtime monitoring, significantly reducing system complexity and operational costs.