Integration of digital twin technology in the predictive maintenance of complex equipment and processes marks a significant advancement in industrial operations. At Bosch, we develop digital twins for turbomachines to predict and mitigate potential failures. While early detection of machine failure through data-driven vibration analysis is a major achievement, identifying the root cause of the impending failures remains challenging, often due to a lack of high-quality data. In this context, understanding causality, the cause-and-effect relationship, provides valuable insights into how interventions on one variable will influence another [1]. In short, causality helps in fixing the root cause rather than working on the symptoms of machine failure. Impacted Key Performance Indicators (KPIs) are mean time between failure, Remaining Useful Lifetime (RUL), operations & maintenance expenses, etc. We used multivariate time series data of a degrading gas turbine engine from sensors monitoring several rotating components such as a fan, compressors and turbines. Using this data, we trained machine learning models achieving >90% accuracy in predicting the RUL of the engine. Although our models effectively predict RUL, they do not reveal the root cause of engine degradation. Traditional methods like feature importance estimation through weight-gain-cover, ordinary least square regression or shapley additive explanations primarily measure correlation rather than causality. These methods assume feature independence and overlook the influence of other features in the causal pathways. To this end, we quantified the causal effect of every feature on the RUL by applying the backdoor criterion, which controls the common-cause features isolating the true causal relationships. Leveraging simulation models and domain knowledge, we constructed causal graphs that depict interactions between engine components and their influence on RUL. We found and validated that the major components of the engine responsible for its degradation are the fan and the high-pressure compressor. We will present the methodology for estimating causal effects, report our results and compare them with different techniques for the digital twin of a gas turbine engine, demonstrating how causality enhances predictive maintenance in industrial applications.