In previous works, the authors have shown the feasibility of using classical and robust principal component analysis for damage detection on structures when ultrasonic guided waves are used. It has also been demonstrated that robust principal component analysis presents a higher probability of detection accuracy when data are corrupted. In the present work, a robust principal component analysis orthogonal distance is proposed as a new feature for damage detection strategy based on ultrasonic guided waves on structures subjected to uniform temperature changes. The effect of this temperature fluctuation on the signal propagation and also in the new feature is analyzed. Temperature compensation is applied to mitigate the effect of temperature changes on the reliability of the damage detection methodology. The proposed feature and damage detection strategy that considers these effects are tested on two structures: a laboratory scale composite plate and a large-scale complex composite that is representative of a component from an aerospace application. The promising result proves the ability of the new feature as a damage detection tool.