A novel fault diagnosis strategy for LED lamps via light output time-frequency characteristics analysis and machine learning.

Faulty LED lamps can cause a decrease in light efficiency, lead to flicker, and have a negative impact on creating a reliable, stable, and healthy light environment. However, many LED lamps' faults are difficult to detect by electrical parameter measurements or naked-eye observation. Consequently, in this paper, a novel fault diagnosis strategy is proposed by analyzing light output time-frequency characteristics of LED lamps. The proposed fault diagnosis strategy contains three stages: (1) collecting the light output signal of LED lamps, (2) extracting the light output time-frequency characteristics of LED lamps by VMD and energy entropy calculation, and (3) employing SVM to construct the fault diagnosis model which used to identify the faulty LED lamps. To validate the feasibility and effectiveness of the proposed fault diagnosis strategy, simulation experiments are conducted, and the light output signals of LED lamps are collected as experiment datasets using the 10 kHz sampling frequency. The results demonstrate that the proposed fault diagnosis strategy can identify faults effectively, and average accuracy rate can reach to over 92%. This study can help promote the development of large-scale LED lamp maintenance management technology, and bring great benefits for the reliable and healthy operation of large-scale LED lamps particularly.