Degradation Process Modeling based on Reliability Test and Machine Learning Regression

A degradation process is the deterioration of an object's internal and external properties, resulting in a decline in performance quality and ability to meet design and operational requirements. Modelling the degradation process has received significant attention from reliability and statistical scientists. Many methods have been proposed and developed to model the degradation process and also to predict and estimate reliability measures. The regression methods are essential methods in machine learning, and it has proved the huge potential for modelling the deterioration process of the objects. In essence, machine learning regression is a concept that represents a series of methods based on i) supervised learning and on collecting data from actual object operations or on ii) reliability tests in laboratories. This paper presents general knowledge of machine-learning regression and uses them to model the degradation process of light-emitting diodes (LEDs) based on the data obtained from reliability tests in laboratories in two specific cases of input data. Afterwards, the paper compares the performance of these methods and assesses the suitability and effectiveness of these methods for modelling the degradation process of LEDs.

Keywords: Light-emitting diode, Degradation process, Machine learning regression, Reliability testing, Support vector regression, Gaussian process regression, k-nearest neighbors, Random forest.

Download PDF