Accelerated Life Test And Life Evaluation Of Semiconductor Lighting

In recent years, along with the Light Emitting Diode(LEDs) development and application in areas such as information display and semiconductor lighting source, the importance of its reliability is also growing. But the existing reliability research of LEDs mostly concentrated in the failure analysis of a single LED module, or applied conventional accelerated life test to forecast and analysis the LEDs reliability. Its reliability needs to be assess fast and accurately due to LEDs long life, fast upgrading speed, high cost and long cycle life test. The evaluation of reliability contributes to guide LED light source system to design structure, improve manufacture process, filter premature products and verify system reliability test.This thesis discusses the accelerated life test of LED light source system from the designing test system and test data life evaluation two aspects. Its mainly work begins with the basis of analyzing the affecting factors of LED life, selecting temperature and electrical stresses as the experimental stress. And ascertaining test parameters, failure criterion and test stress range. Afterwards, this thesis presents a reliability assessment method, based on acceleration performance degradation of the LED light source. Includes determining the different stress levels of degenerate trajectory equation; In addition, according to a given accelerated degradation test threshold of the luminous flux, predicting pseudo failure life and estimate the optimal distribution model under each group stress test samples condition; Next, founding an accelerated model construction of temperature and electric stress, and extrapolating LED life parameters and other reliability under normal conditions. Finally, in order to reduce life evaluation error futher, the thesis put forward a kind of life evalution method based on ant colony neural network. The results show that the prediction error smaller and faster. The life prediction model has a certain universality, and it's generalize to apply reliability prediction and evaluationin other electronic components.