Turbofan engines are an important part of aircraft that generate a large amount of multivariate sensor data during operation. The C-MAPSS FD001 dataset released by NASA accurately reflects turbofan engine degradation data. FD001 contains rows of data representing one operational cycle of a specific engine with information about the Engine ID, Cycle, operational settings, and 21 sensor readings that reflect the physical condition and performance of the engine, but these are not labeled, making it difficult to identify anomalies directly. Therefore, an automated method is needed that can learn normal patterns and identify data deviations accurately. This study developed an autoencoder model based on Mean Squared Error (MSE) reconstruction to find anomalies in the C-MAPSS FD001 dataset. This model was created with 15 inputs, 8 latent spaces, 15 inputs architecture and trained using 20,631 normalized training data (train_FD001). Anomalies were determined based on a 95th percentile threshold of the MSE value distribution. Testing was performed on test data (test_FD001) using a similar process. Testing results on 13,096 test data points showed an average MSE value of 0.001823 with a standard deviation of 0.001032. From the 95th percentile threshold value of 0,003785, 131 data points, or about ±1%, were identified as anomalies. The low MSE value for most of the data indicates that the model can reconstruct normal data patterns well, while data with high MSE values can be identified as anomalies. This study confirms that autoencoders with error reconstruction are effective for detecting anomalies in unlabeled turbofan engine sensor data.