The Skin Air Outlet valve(SAO) is a critical component of the aircraft’s air circulation system. Located outside the avionics bay, it periodically opens and closes to regulate air circulation, preventing electronic devices from overheating and ensuring the stability of the aircraft cabin pressure. Currently, fault detection of the SAO is based on fixed time thresholds according to the sensor state changes, which has limitations in maintenance efficiency and aircraft safety. This study proposes a data-driven approach for constructing baseline model and health index(HI) for the SAO of the Airbus A330. A substantial dataset related to the SAO was built using Quick Access Recorder(QAR) data, which includes both normal and faulty states. To explore the extensive QAR data of the SAO system, a baseline constructing approach based on deep learning was proposed. For effective processing of high-dimensional feature data of the SAO and modeling of the SAO baseline, the Two-Stage Attention(TSA) mechanism-based algorithm is proposed. Additionally, this study proposes a method for constructing HI based on the SAO baseline, along with experiments on the diagnosis and prognosis of the SAO based on the HI. Compared to the conventional method, the proposed HI-based fault diagnosis and prognosis method demonstrated excellent diagnostic effectiveness and revealed a clearer trend of health degradation in the SAO.