TY - JOUR

T1 - KBRDBN

T2 - An Interpretable Deep Belief Network for the Fault Diagnosis of the Trolley Mechanism in Ship-to-Shore Cranes

AU - Liao, Xiaoqiang

AU - Ming, Xinguo

AU - Xia, Min

PY - 2023/9/25

Y1 - 2023/9/25

N2 - The fault diagnosis of the trolley mechanism in Ship-to-Shore Cranes (STSC) plays a critical role in guaranteeing the stability and safety of the container handling at the port. The existing models based on Deep Neural Networks (DNNs) have achieved slight success in the fault diagnosis of the trolley mechanism in the STSC system. Nevertheless, due to the black box characteristics of DNNs, these diagnostic models cannot provide a reasonable explanation for their decisions. Based on the predictions made by DNNs, it is yet challenging for experts to make informed and trustworthy decisions. To address this, this paper proposes a Knowledge-Based Reverse Deep Belief Network (KBRDBN) to build an efficient neural-symbolic system capable of interpreting feature learning and reasoning of DNNs. In KBRDBN, two types of relational knowledge (i.e., confidence and soft rules) are extracted. Confidence rules can explain how the Deep Belief Network works and produce a latent feature for raw vibration signals of the trolley mechanism. Soft rules can denote inherent uncertainty and conduct quantitative reasoning for the diagnostic decisions of the trolley mechanism. The experimental study on a testbed demonstrates the remarkable performance of the proposed approach in interpretability, uncertainty-handling capability, and fault recognition of the trolley mechanism.

AB - The fault diagnosis of the trolley mechanism in Ship-to-Shore Cranes (STSC) plays a critical role in guaranteeing the stability and safety of the container handling at the port. The existing models based on Deep Neural Networks (DNNs) have achieved slight success in the fault diagnosis of the trolley mechanism in the STSC system. Nevertheless, due to the black box characteristics of DNNs, these diagnostic models cannot provide a reasonable explanation for their decisions. Based on the predictions made by DNNs, it is yet challenging for experts to make informed and trustworthy decisions. To address this, this paper proposes a Knowledge-Based Reverse Deep Belief Network (KBRDBN) to build an efficient neural-symbolic system capable of interpreting feature learning and reasoning of DNNs. In KBRDBN, two types of relational knowledge (i.e., confidence and soft rules) are extracted. Confidence rules can explain how the Deep Belief Network works and produce a latent feature for raw vibration signals of the trolley mechanism. Soft rules can denote inherent uncertainty and conduct quantitative reasoning for the diagnostic decisions of the trolley mechanism. The experimental study on a testbed demonstrates the remarkable performance of the proposed approach in interpretability, uncertainty-handling capability, and fault recognition of the trolley mechanism.

KW - Electrical and Electronic Engineering

KW - Instrumentation

U2 - 10.1109/tim.2023.3318717

DO - 10.1109/tim.2023.3318717

M3 - Journal article

JO - IEEE Transactions on Instrumentation and Measurement

JF - IEEE Transactions on Instrumentation and Measurement

SN - 0018-9456

ER -