Quantum mechanics

Process (computing)

Engineering

Transient (computer programming)

Reliability engineering

Aerospace engineering

Materials science

Shock (circulatory)

Automotive industry

Operating system

Power (physics)

Component (thermodynamics)

Computer science

Soldering

Thermodynamics

Physics

Thermal

Internal medicine

Automotive engineering

Reliability (semiconductor)

Artificial intelligence

Composite material

Medicine

Meteorology

Artificial neural network

Light-emitting diode

Electrical engineering

Accurately predicting the lifetime of an electrical component is an important part of the design process, both in the initial creation of the component and later when integrating it into larger devices. This is especially important in applications with high reliability standards, such as automotive LEDs. In this work, we compare methods to predict failures in solder joints of LEDs during accelerated stress testing (temperature shock tests) using transient thermal analysis (TTA). We compare a statistical approach and the use of artificial neural networks (ANN) with memory for two types of prediction: the state at a specific number of shock cycle and for a fixed number of cycles into the future. Our results show that the data-driven approach improves the prediction compared to classical methods in both scenarios.

Comparing prediction methods for LED failure measured with Transient Thermal Analysis