Semi-Supervised Learning-Driven LED Lamp Fault Diagnosis Method Using Coefficient of Variation-Based Light Output Analysis

Quality inspection is a critical step in the LED lighting manufacturing process. However, reliable and efficient fault diagnosis methods are still difficult to provide. To address this issue, this paper proposes a fault diagnosis method for LED lamps based on an unsupervised anomaly detection algorithm and semi-supervised learning-driven classification. First, a coefficient of variation in light output signal is introduced to describe the operational status of the LED lamps. Second, an improved unsupervised LOF algorithm is employed to filter anomalous data, reducing the interference of potential outliers during data acquisition. Third, the semi-supervised algorithm is employed for feature analysis and classification, achieving an efficient and reliable LED fault diagnosis system. In experiments, the fault diagnosis accuracy for OSRAM 8.5W LED lamps reached 96.58%, confirming the effectiveness of the proposed method.