Diagnosability-optimized design of unmanned aerial vehicles based on structural analysis and maximum mean covariance differences

This study presents a proposed strategy for optimizing diagnosability in unmanned aerial vehicles (UAVs) through a combination of Structural Analysis (SA) and Maximum Mean Covariance Difference (MMCD). Initially, diagnosability is qualitatively assessed using SA. Residuals are then designed based on the minimal structural overdetermination (MSO) using the sequential residual generation approach. Subsequently, diagnosability is reframed as a measurement of residual variance in the probability distribution. To address the absence of Maximum Mean Difference (MMD) in covariance difference measurement, Maximum Covariance Difference (MCD) is introduced. Furthermore, a method for quantitatively evaluating diagnosability based on MMCD is proposed, incorporating both MMD and MCD. Finally, an optimization model for diagnosability is established to ensure the efficient completion of fault diagnostic tasks with minimal cost and maximum effectiveness. The proposed method's rationality and effectiveness are validated through simulation using the Tree Seed Algorithm (TSA) on a fixed-wing UAV model.