Abstract: The fault diagnosis of aero-engines was confronted with a data skew issue, where the number of fault samples was sig-nificantly fewer than normal samples, and the fault samples couldn"t adequately represent the entire operating conditions, resulting in poor generalization ability of conventional classification models. To address this issue, an improved deep support vector data descrip-tion-based time series anomaly detection model was proposed. Initially, a Long Short-Term Memory Recurrent Neural Network (LSTM) was employed to map the inputs and outputs of samples, forming temporal anomaly vectors with actual collected outputs. Subsequently, deep support vector data description (SVDD) incorporating Variational Autoencoder (VAE) was utilized to achieve anomaly detection for aero-engine time series data. Experimental results on a certain type of aero-engine ground test platform demonstrated that compared to contrastive methods such as Isolation Forest (IF), Transformer-based Anomaly Detection Model (TranAD), and GANomaly, the proposed algorithm achieved an Area Under the Curve (AUC) value of 0.9878, indicating su-perior anomaly detection performance. Therefore, the proposed algorithm can effectively be applied to various anomaly detection and fault diagnosis tasks in aero-engine systems.