Various embodiments described herein relate to an anomaly detection framework adaptable to different asset types. In this regard, a data stream associated with a first asset is received. The data stream is then processed to generate output data by encoding the data stream based on historical data associated with the first asset, the historical data comprising clustered data representative of fault states and one or more non-fault states. Furthermore, in accordance with a determination that the generated output data is indicative of a potential fault of the first asset, fault data indicative of the potential fault is generated and caused to be transmitted to an administrative device for display.