Aiming at the challenges of multi-source heterogeneous data fusion, large working condition disturbance and high real-time requirement in fault diagnosis of ship lock electrical control system, this paper proposes an adaptive fault diagnosis framework driven by multi-source data. The framework constructs a dynamic weight allocation mechanism through information entropy and fuzzy logic to solve the problems of asynchronous data alignment and credibility difference. Design 1D-CNN and graph-volume-product dual-branch network to extract time series and spatial features respectively, so as to improve the ability of compound fault identification. The incremental updating strategy of experience playback and knowledge distillation is introduced to realize the continuous adaptation and lightweight deployment of the model under the aging equipment and changing working conditions. The experiment is based on the 12-month operation data of the real ship lock, covering typical faults such as motor overload, bearing wear, transmission chain sticking and so on. The results show that the average accuracy rate of the proposed method is 89.8%, the false alarm rate is less than 4.5%, the detection delay is only 1.8s, and the model parameters are controlled within 423 thousand, which meets the requirements of PLC edge deployment. The research effectively alleviates the diagnosis bottleneck under the constraints of "strong interference, low computational power and high safety", and provides technical support for the reliable operation and maintenance of the smart ship lock.
Biao Li (Sun,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: