Hierarchical Duplex Diversity Entropy for Diesel Injector fault diagnosis

In the realm of diesel fuel injector condition monitoring and fault diagnosis, extracting nonlinear dynamic fault features is a crucial scientific challenge. Multiscale Diversity Entropy (MDE) is a novel approach for measuring the dynamics of nonlinear signals. However, due to the limitations of its similarity definition and the lack of high-frequency information, it leads to the loss of deep information, thereby affecting the accuracy of fault diagnosis. Aiming at the intrinsic shortcomings of MDE, this paper proposes a novel method termed Hierarchical Duplex Diversity Entropy (HDDE). When calculating trajectory diversity, HDDE introduces the resemblance between adjacent trajectories and the next nearest trajectories in the space to characterize the complexity of the system, thereby enabling interaction between different types of trajectories. At the same time, HDDE combines hierarchical analysis to extract high-frequency and low-frequency features at multiple levels, effectively solving the problem of fault information loss. Based on the advantages of HDDE, this paper proposes a fuel injector fault diagnosis framework combining HDDE and Least Squares Support Vector Machine optimized by Improved Grey Wolf Optimizer (IGWO-LSSVM). We applied the proposed method to test data of different fault types and different fault severity levels of fuel injectors. The results show that HDDE is better than MBDE and HDE in reflecting the dynamic changes of fuel injector fault signals and demonstrates the best feature extraction performance. Moreover, when contrasted with current approaches, the method introduced in this paper enhances the precision and reliability of fault detection, thereby offering a superior technical solution for diagnosing faults in diesel engine fuel injectors.