Research on Fault Diagnosis of Six-Phase Propulsion Motor Drive Inverter for Marine Electric Propulsion System Based on Res-BiLSTM

Abstract

To ensure the implementation of the marine electric propulsion self-healing strategy after faults, it is necessary to diagnose and accurately classify the faults. Considering the characteristics of the residual network (ResNet) and bidirectional long short-term memory (BiLSTM), the Res-BiLSTM deep learning algorithm is used to establish a fault diagnosis model to distinguish the types of electric drive faults. First, the powerful fault feature extraction ability of the residual network is used to deeply mine the fault features in the signals. Then, perform time-series learning through a bidirectional long short-term memory network, and further excavate the transient time-series features in the fault features so as to achieve the accurate classification of drive inverter faults. The effectiveness of the method is verified using noise-free fault data, and the robustness of the method is verified using data with varying degrees of noise. The results show that compared with conventional deep learning algorithms, Res-BiLSTM has the fastest and most stable training process, the diagnostic performance is improved, and the accuracy can be maintained over 95% under 25–19 dB. It has certain robustness and can be applied to marine electric propulsion systems drive inverter fault diagnosis, and its results can provide data support for the implementation of self-healing control strategies.

A novel approach for bearings multiclass fault diagnosis fusing multiscale deep convolution and hybrid attention networks

“…The proposed method is validated using the CWRU dataset and compared with five other fault diagnosis methods, namely, MCNN, WDCNN [33], ResNet [34], DRSN [35], RES-BiLSTM [36] and DRSN-BiLSTM [37]. This comparison test is performed on a computer with an i7-9700F CPU and GeForce RTX 2080 GPU using TensorFlow2.6 software.…”

A Fault-Tolerant Control Method Based on Reconfiguration SPWM Signal for Cascaded Multilevel IGBT-Based Propulsion in Electric Ships

“…For fault diagnosis in a ship's DC electrical system, there has been a lot of research developments, such as a convolutional-neural-network-based method [21], Res-BiLSTM [22], and a layering linear discriminant analysis [23].…”

“…Fault diagnosis is an essential step that is responsible for determining fault information and activating the corresponding fault-tolerant control method. Some effective diagnosis methods have been proposed in the field of ships [21][22][23]. This paper mainly focuses on the research on fault-tolerant control of open-circuit faults.…”

Design and Analysis Models With PID and PID Fuzzy Controllers for Six-Phase Drive

“…The multiphase motors and drives with redundant phase number characterized by high reliability: they are able successfully but with greater losses operate under fault conditions up to single winding remains, but they are not able to start without three windings supplied. [10][11][12][13] The motor dynamic performance is describing by state-space equations.…”

“…It is possible to achieve smaller size and weight of electric drives and their control electronics.The multiphase motors and drives with redundant phase number characterized by high reliability: they are able successfully but with greater losses operate under fault conditions up to single winding remains, but they are not able to start without three windings supplied. [10][11][12][13] The motor dynamic performance is describing by state-space equations.Solution of those gives information about dynamic and steadystate characteristics of six-phase motor drive. It can be used to analyse behaviour of the drive and design control methods improving performance.…”