Wearable devices are increasingly used for personal health monitoring. However, existing decision-making approaches often raise concerns related to outcome interpretability, context awareness and data privacy. This work proposes a novel knowledge-based framework for on-board health monitoring and inference on Apple Watch devices. HealthKit data are annotated as Description Logic concept expressions with respect to a reference ontology to produce a dynamic Personal Health Knowledge Graph. Then, by applying semantic inferences via an embedded reasoner, the framework enables local, explainable analysis of user health status without transmitting sensitive data to external devices or services. A case study focusing on asthma monitoring is presented, in which the severity of symptoms is estimated through locally available data and a custom ontology aligned with the Asthma Control Questionnaire (ACQ) clinical gold standard. A prototypical stand-alone watchOS application demonstrates the feasibility of the approach.