Fault Diagnosis Strategy for Engine Oil Systems in Field Testing

To address the ineffective fault isolation of static diagnosis strategies under field-testing constraints, this paper proposes a dynamic fault diagnosis approach for oil systems based on an improved information entropy algorithm, along with a multifault isolation method. The study analyzes field constraints, introduces the algorithm, and describes a diagnostic framework adaptable to imperfect testing. Using a temporary local fault–test correlation matrix (D-matrix) enhanced with field constraints, dynamic strategy formulation and multifault isolation are demonstrated. Experimental results showed significant improvements over rollout and particle swarm optimization (PSO) algorithms: the isolation rate increased by about 17% and 14%, the average misdiagnosis cost dropped by about 63% and 41%, the detection rate rose by 8.4% and 6.9%, the type I false alarm rate decreased by 97%, and runtime was reduced by 73% and 83%. In multifault scenarios, correctly isolated samples increased from 33 to 40 for two faults and from 70 to 80 for three faults, confirming the method’s effectiveness in improving fault isolation for field applications.